Download Full Neural Information Processing: 25th International Conference, ICONIP 2018, Siem Reap, Cambodia, December 13–16, 2018, Proceedings, Part II Long Cheng PDF All Chapters

Download the Full Version of textbook for Fast Typing at textbookfull.
com
Neural Information Processing: 25th International

Conference, ICONIP 2018, Siem Reap, Cambodia,
December 13–16, 2018, Proceedings, Part II Long
Cheng
https://textbookfull.com/product/neural-information-
processing-25th-international-conference-iconip-2018-siem-
reap-cambodia-december-13-16-2018-proceedings-part-ii-long-
cheng/
OR CLICK BUTTON
DOWNLOAD NOW
Download More textbook Instantly Today - Get Yours Now at textbookfull.com

Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.

Conference, ICONIP 2018, Siem Reap, Cambodia, December
13–16, 2018, Proceedings, Part II Long Cheng
https://textbookfull.com/product/neural-information-processing-25th-
international-conference-iconip-2018-siem-reap-cambodia-
december-13-16-2018-proceedings-part-ii-long-cheng/
textboxfull.com

13–16, 2018, Proceedings, Part III Long Cheng
december-13-16-2018-proceedings-part-iii-long-cheng/
textboxfull.com

13–16, 2018, Proceedings, Part V Long Cheng
december-13-16-2018-proceedings-part-v-long-cheng/
textboxfull.com

13–16, 2018, Proceedings, Part VI Long Cheng
december-13-16-2018-proceedings-part-vi-long-cheng/
textboxfull.com
13–16, 2018, Proceedings, Part VII Long Cheng
december-13-16-2018-proceedings-part-vii-long-cheng/
textboxfull.com
Neural Information Processing 27th International

Conference ICONIP 2020 Bangkok Thailand November 23 27
2020 Proceedings Part II Haiqin Yang
international-conference-iconip-2020-bangkok-thailand-
november-23-27-2020-proceedings-part-ii-haiqin-yang/
textboxfull.com
Computational Science ICCS 2018 18th International

Conference Wuxi China June 11 13 2018 Proceedings Part II
Yong Shi
https://textbookfull.com/product/computational-science-iccs-2018-18th-
international-conference-wuxi-china-june-11-13-2018-proceedings-part-
ii-yong-shi/
textboxfull.com

2020 Proceedings Part IV Haiqin Yang
november-18-22-2020-proceedings-part-iv-haiqin-yang/
textboxfull.com

2020 Proceedings Part V Haiqin Yang
november-18-22-2020-proceedings-part-v-haiqin-yang/
textboxfull.com
Long Cheng
Andrew Chi Sing Leung
Seiichi Ozawa (Eds.)
LNCS 11302
Neural
Information Processing
25th International Conference, ICONIP 2018
Siem Reap, Cambodia, December 13–16, 2018
Proceedings, Part II
123
Lecture Notes in Computer Science 11302
Commenced Publication in 1973
Founding and Former Series Editors:
Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen
Editorial Board
David Hutchison
Lancaster University, Lancaster, UK
Takeo Kanade
Carnegie Mellon University, Pittsburgh, PA, USA
Josef Kittler
University of Surrey, Guildford, UK
Jon M. Kleinberg
Cornell University, Ithaca, NY, USA
Friedemann Mattern
ETH Zurich, Zurich, Switzerland
John C. Mitchell
Stanford University, Stanford, CA, USA
Moni Naor
Weizmann Institute of Science, Rehovot, Israel
C. Pandu Rangan
Indian Institute of Technology Madras, Chennai, India
Bernhard Steffen
TU Dortmund University, Dortmund, Germany
Demetri Terzopoulos
University of California, Los Angeles, CA, USA
Doug Tygar
University of California, Berkeley, CA, USA
Gerhard Weikum
Max Planck Institute for Informatics, Saarbrücken, Germany
More information about this series at http://www.springer.com/series/7407
Long Cheng Andrew Chi Sing Leung
•
Seiichi Ozawa (Eds.)
Neural
Information Processing
25th International Conference, ICONIP 2018
Siem Reap, Cambodia, December 13–16, 2018
Proceedings, Part II
123
Editors
Long Cheng Seiichi Ozawa
The Chinese Academy of Sciences Kobe University
Beijing, China Kobe, Japan
City University of Hong Kong
Kowloon, Hong Kong SAR, China
ISSN 0302-9743 ISSN 1611-3349 (electronic)

Lecture Notes in Computer Science
ISBN 978-3-030-04178-6 ISBN 978-3-030-04179-3 (eBook)
https://doi.org/10.1007/978-3-030-04179-3
Library of Congress Control Number: 2018960916
LNCS Sublibrary: SL1 – Theoretical Computer Science and General Issues
© Springer Nature Switzerland AG 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now
known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, express or implied, with respect to the material contained herein or for any errors or
omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
The 25th International Conference on Neural Information Processing (ICONIP 2018),

the annual conference of the Asia Pacific Neural Network Society (APNNS), was held
in Siem Reap, Cambodia, during December 13–16, 2018. The ICONIP conference
series started in 1994 in Seoul, which has now become a well-established and
high-quality conference on neural networks around the world. Siem Reap is a gateway
to Angkor Wat, which is one of the most important archaeological sites in Southeast
Asia, the largest religious monument in the world. All participants of ICONIP 2018 had
a technically rewarding experience as well as a memorable stay in this great city.
In recent years, the neural network has been significantly advanced with the great
developments in neuroscience, computer science, cognitive science, and engineering.
Many novel neural information processing techniques have been proposed as the
solutions to complex, networked, and information-rich intelligent systems. To dis-
seminate new findings, ICONIP 2018 provided a high-level international forum for
scientists, engineers, and educators to present the state of the art of research and
applications in all fields regarding neural networks.
With the growing popularity of neural networks in recent years, we have witnessed
an increase in the number of submissions and in the quality of submissions. ICONIP
2018 received 575 submissions from 51 countries and regions across six continents.
Based on a rigorous peer-review process, where each submission was reviewed by at
least three experts, a total of 401 high-quality papers were selected for publication in
the prestigious Springer series of Lecture Notes in Computer Science. The selected
papers cover a wide range of subjects that address the emerging topics of theoretical
research, empirical studies, and applications of neural information processing tech-
niques across different domains.
In addition to the contributed papers, the ICONIP 2018 technical program also
featured three plenary talks and two invited talks delivered by world-renowned
scholars: Prof. Masashi Sugiyama (University of Tokyo and RIKEN Center for
Advanced Intelligence Project), Prof. Marios M. Polycarpou (University of Cyprus),
Prof. Qing-Long Han (Swinburne University of Technology), Prof. Cesare Alippi
(Polytechnic of Milan), and Nikola K. Kasabov (Auckland University of Technology).
We would like to extend our sincere gratitude to all members of the ICONIP 2018
Advisory Committee for their support, the APNNS Governing Board for their guid-
ance, the International Neural Network Society and Japanese Neural Network Society
for their technical co-sponsorship, and all members of the Organizing Committee for all
their great effort and time in organizing such an event. We would also like to take this
opportunity to thank all the Technical Program Committee members and reviewers for
their professional reviews that guaranteed the high quality of the conference pro-
ceedings. Furthermore, we would like to thank the publisher, Springer, for their
sponsorship and cooperation in publishing the conference proceedings in seven vol-
umes of Lecture Notes in Computer Science. Finally, we would like to thank all the
VI Preface
speakers, authors, reviewers, volunteers, and participants for their contribution and
support in making ICONIP 2018 a successful event.
October 2018 Jun Wang

Long Cheng
Seiichi Ozawa
ICONIP 2018 Organization
General Chair
Jun Wang City University of Hong Kong,
Hong Kong SAR, China
Advisory Chairs
Akira Hirose University of Tokyo, Tokyo, Japan
Soo-Young Lee Korea Advanced Institute of Science and Technology,
South Korea
Derong Liu Institute of Automation, Chinese Academy of Sciences,
China
Nikhil R. Pal Indian Statistics Institute, India
Program Chairs
Long Cheng Institute of Automation, Chinese Academy of Sciences,
China
Andrew C. S. Leung City University of Hong Kong, Hong Kong SAR,
China
Seiichi Ozawa Kobe University, Japan
Special Sessions Chairs

Shukai Duan Southwest University, China
Kazushi Ikeda Nara Institute of Science and Technology, Japan
Qinglai Wei Institute of Automation, Chinese Academy of Sciences,
China
Hiroshi Yamakawa Dwango Co. Ltd., Japan
Zhihui Zhan South China University of Technology, China
Tutorial Chairs
Hiroaki Gomi NTT Communication Science Laboratories, Japan
Takashi Morie Kyushu Institute of Technology, Japan
Kay Chen Tan City University of Hong Kong, Hong Kong SAR,
China
Dongbin Zhao Institute of Automation, Chinese Academy of Sciences,
China
VIII ICONIP 2018 Organization
Publicity Chairs
Zeng-Guang Hou Institute of Automation, Chinese Academy of Sciences,
China
Tingwen Huang Texas A&M University at Qatar, Qatar
Chia-Feng Juang National Chung-Hsing University, Taiwan
Tomohiro Shibata Kyushu Institute of Technology, Japan
Publication Chairs
Xinyi Le Shanghai Jiao Tong University, China
Sitian Qin Harbin Institute of Technology Weihai, China
Zheng Yan University Technology Sydney, Australia
Shaofu Yang Southeast University, China
Registration Chairs
Shenshen Gu Shanghai University, China
Qingshan Liu Southeast University, China
Ka Chun Wong City University of Hong Kong,
Conference Secretariat
Ying Qu Dalian University of Technology, China
Program Committee
Hussein Abbass University of New South Wales at Canberra, Australia
Choon Ki Ahn Korea University, South Korea
Igor Aizenberg Texas A&M University at Texarkana, USA
Shotaro Akaho National Institute of Advanced Industrial Science
and Technology, Japan
Abdulrazak Alhababi UNIMAS, Malaysia
Cecilio Angulo Universitat Politècnica de Catalunya, Spain
Sabri Arik Istanbul University, Turkey
Mubasher Baig National University of Computer and Emerging
Sciences Lahore, India
Sang-Woo Ban Dongguk University, South Korea
Tao Ban National Institute of Information and Communications
Technology, Japan
Boris Bačić Auckland University of Technology, New Zealand
Xu Bin Northwestern Polytechnical University, China
David Bong Universiti Malaysia Sarawak, Malaysia
Salim Bouzerdoum University of Wollongong, Australia
Ivo Bukovsky Czech Technical University, Czech Republic
ICONIP 2018 Organization IX
Ke-Cai Cao Nanjing University of Posts and Telecommunications,

China
Elisa Capecci Auckland University of Technology, New Zealand
Rapeeporn Chamchong Mahasarakham University, Thailand
Jonathan Chan King Mongkut’s University of Technology Thonburi,
Thailand
Rosa Chan City University of Hong Kong, Hong Kong SAR,
China
Guoqing Chao East China Normal University, China
He Chen Nankai University, China
Mou Chen Nanjing University of Aeronautics and Astronautics,
China
Qiong Chen South China University of Technology, China
Wei-Neng Chen Sun Yat-Sen University, China
Xiaofeng Chen Chongqing Jiaotong University, China
Ziran Chen Bohai University, China
Jian Cheng Chinese Academy of Sciences, China
Long Cheng Chinese Academy of Sciences, China
Wu Chengwei Bohai University, China
Zheru Chi The Hong Kong Polytechnic University, SAR China
Sung-Bae Cho Yonsei University, South Korea
Heeyoul Choi Handong Global University, South Korea
Hyunsoek Choi Kyungpook National University, South Korea
Supannada Chotipant King Mongkut’s Institute of Technology Ladkrabang,
Thailand
Fengyu Cong Dalian University of Technology, China
Jose Alfredo Ferreira Costa Federal University of Rio Grande do Norte, Brazil
Ruxandra Liana Costea Polytechnic University of Bucharest, Romania
Jean-Francois Couchot University of Franche-Comté, France
Raphaël Couturier University of Bourgogne Franche-Comté, France
Jisheng Dai Jiangsu University, China
Justin Dauwels Massachusetts Institute of Technology, USA
Dehua Zhang Chinese Academy of Sciences, China
Mingcong Deng Tokyo University of Agriculture and Technology,
Japan
Zhaohong Deng Jiangnan University, China
Jing Dong Chinese Academy of Sciences, China
Qiulei Dong Chinese Academy of Sciences, China
Kenji Doya Okinawa Institute of Science and Technology, Japan
El-Sayed El-Alfy King Fahd University of Petroleum and Minerals,
Saudi Arabia
Mark Elshaw Nottingham Trent International College, UK
Peter Erdi Kalamazoo College, USA
Josafath Israel Espinosa Auckland University of Technology, New Zealand
Ramos
Issam Falih Paris 13 University, France
X ICONIP 2018 Organization
Bo Fan Zhejiang University, China

Yunsheng Fan Dalian Maritime University, China
Hao Fang Beijing Institute of Technology, China
Jinchao Feng Beijing University of Technology, China
Francesco Ferracuti Università Politecnica delle Marche, Italy
Chun Che Fung Murdoch University, Australia
Wai-Keung Fung Robert Gordon University, UK
Tetsuo Furukawa Kyushu Institute of Technology, Japan
Hao Gao Nanjing University of Posts and Telecommunications,
China
Yabin Gao Harbin Institute of Technology, China
Yongsheng Gao Griffith University, Australia
Tom Gedeon Australian National University, Australia
Ong Sing Goh Universiti Teknikal Malaysia Melaka, Malaysia
Iqbal Gondal Federation University Australia, Australia
Yue-Jiao Gong Sun Yat-sen University, China
Shenshen Gu Shanghai University, China
Chengan Guo Dalian University of Technology, China
Ping Guo Beijing Normal University, China
Shanqing Guo Shandong University, China
Xiang-Gui Guo University of Science and Technology Beijing, China
Zhishan Guo University of Central Florida, USA
Christophe Guyeux University of Franche-Comte, France
Masafumi Hagiwara Keio University, Japan
Saman Halgamuge The University of Melbourne, Australia
Tomoki Hamagami Yokohama National University, Japan
Cheol Han Korea University at Sejong, South Korea
Min Han Dalian University of Technology, China
Takako Hashimoto Chiba University of Commerce, Japan
Toshiharu Hatanaka Osaka University, Japan
Wei He University of Science and Technology Beijing, China
Xing He Southwest University, China
Xiuyu He University of Science and Technology Beijing, China
Akira Hirose The University of Tokyo, Japan
Daniel Ho City University of Hong Kong, Hong Kong SAR,
China
Katsuhiro Honda Osaka Prefecture University, Japan
Hongyi Li Bohai University, China
Kazuhiro Hotta Meijo University, Japan
Jin Hu Chongqing Jiaotong University, China
Jinglu Hu Waseda University, Japan
Xiaofang Hu Southwest University, China
Xiaolin Hu Tsinghua University, China
He Huang Soochow University, China
Kaizhu Huang Xi’an Jiaotong-Liverpool University, China
Long-Ting Huang Wuhan University of Technology, China
ICONIP 2018 Organization XI
Panfeng Huang Northwestern Polytechnical University, China

Tingwen Huang Texas A&M University, USA
Hitoshi Iima Kyoto Institute of Technology, Japan
Kazushi Ikeda Nara Institute of Science and Technology, Japan
Hayashi Isao Kansai University, Japan
Teijiro Isokawa University of Hyogo, Japan
Piyasak Jeatrakul Mae Fah Luang University, Thailand
Jin-Tsong Jeng National Formosa University, Taiwan
Sungmoon Jeong Kyungpook National University Hospital, South Korea
Danchi Jiang University of Tasmania, Australia
Min Jiang Xiamen University, China
Yizhang Jiang Jiangnan University, China
Xuguo Jiao Zhejiang University, China
Keisuke Kameyama University of Tsukuba, Japan
Shunshoku Kanae Junshin Gakuen University, Japan
Hamid Reza Karimi Politecnico di Milano, Italy
Nikola Kasabov Auckland University of Technology, New Zealand
Abbas Khosravi Deakin University, Australia
Rhee Man Kil Sungkyunkwan University, South Korea
Daeeun Kim Yonsei University, South Korea
Sangwook Kim Kobe University, Japan
Lai Kin Tunku Abdul Rahman University, Malaysia
Irwin King The Chinese University of Hong Kong,
Yasuharu Koike Tokyo Institute of Technology, Japan
Ven Jyn Kok National University of Malaysia, Malaysia
Ghosh Kuntal Indian Statistical Institute, India
Shuichi Kurogi Kyushu Institute of Technology, Japan
Susumu Kuroyanagi Nagoya Institute of Technology, Japan
James Kwok The Hong Kong University of Science
and Technology, SAR China
Edmund Lai Auckland University of Technology, New Zealand
Kittichai Lavangnananda King Mongkut’s University of Technology Thonburi,
Thailand
Xinyi Le Shanghai Jiao Tong University, China
Minho Lee Kyungpook National University, South Korea
Nung Kion Lee University Malaysia Sarawak, Malaysia
Andrew C. S. Leung City University of Hong Kong, Hong Kong SAR,
China
Baoquan Li Tianjin Polytechnic University, China
Chengdong Li Shandong Jianzhu University, China
Chuandong Li Southwest University, China
Dazi Li Beijing University of Chemical Technology, China
Li Li Tsinghua University, China
Shengquan Li Yangzhou University, China
XII ICONIP 2018 Organization
Ya Li Institute of Automation, Chinese Academy of Sciences,

China
Yanan Li University of Sussex, UK
Yongming Li Liaoning University of Technology, China
Yuankai Li University of Science and Technology of China, China
Jie Lian Dalian University of Technology, China
Hualou Liang Drexel University, USA
Jinling Liang Southeast University, China
Xiao Liang Nankai University, China
Alan Wee-Chung Liew Griffith University, Australia
Honghai Liu University of Portsmouth, UK
Huaping Liu Tsinghua University, China
Huawen Liu University of Texas at San Antonio, USA
Jing Liu Chinese Academy of Sciences, China
Ju Liu Shandong University, China
Qingshan Liu Huazhong University of Science and Technology,
China
Weifeng Liu China University of Petroleum, China
Weiqiang Liu Nanjing University of Aeronautics and Astronautics,
China
Dome Lohpetch King Mongkut’s University of Technology
North Bangoko, Thailand
Hongtao Lu Shanghai Jiao Tong University, China
Wenlian Lu Fudan University, China
Yao Lu Beijing Institute of Technology, China
Jinwen Ma Peking University, China
Qianli Ma South China University of Technology, China
Sanparith Marukatat Thailand’s National Electronics and Computer
Technology Center, Thailand
Tomasz Maszczyk Nanyang Technological University, Singapore
Basarab Matei LIPN Paris Nord University, France
Takashi Matsubara Kobe University, Japan
Nobuyuki Matsui University of Hyogo, Japan
P. Meesad King Mongkut’s University of Technology
North Bangkok, Thailand
Gaofeng Meng Chinese Academy of Sciences, China
Daisuke Miyamoto University of Tokyo, Japan
Kazuteru Miyazaki National Institution for Academic Degrees and Quality
Enhancement of Higher Education, Japan
Seiji Miyoshi Kansai University, Japan
J. Manuel Moreno Universitat Politècnica de Catalunya, Spain
Naoki Mori Osaka Prefecture University, Japan
Yoshitaka Morimura Kyoto University, Japan
Chaoxu Mu Tianjin University, China
Kazuyuki Murase University of Fukui, Japan
Jun Nishii Yamaguchi University, Japan
ICONIP 2018 Organization XIII
Haruhiko Nishimura University of Hyogo, Japan

Grozavu Nistor Paris 13 University, France
Yamaguchi Nobuhiko Saga University, Japan
Stavros Ntalampiras University of Milan, Italy
Takashi Omori Tamagawa University, Japan
Toshiaki Omori Kobe University, Japan
Seiichi Ozawa Kobe University, Japan
Yingnan Pan Northeastern University, China
Yunpeng Pan JD Research Labs, China
Lie Meng Pang Universiti Malaysia Sarawak, Malaysia
Shaoning Pang Unitec Institute of Technology, New Zealand
Hyeyoung Park Kyungpook National University, South Korea
Hyung-Min Park Sogang University, South Korea
Seong-Bae Park Kyungpook National University, South Korea
Kitsuchart Pasupa King Mongkut’s Institute of Technology Ladkrabang,
Thailand
Yong Peng Hangzhou Dianzi University, China
Somnuk Phon-Amnuaisuk Universiti Teknologi Brunei, Brunei
Lukas Pichl International Christian University, Japan
Geong Sen Poh National University of Singapore, Singapore
Mahardhika Pratama Nanyang Technological University, Singapore
Emanuele Principi Università Politecnica elle Marche, Italy
Dianwei Qian North China Electric Power University, China
Jiahu Qin University of Science and Technology of China, China
Sitian Qin Harbin Institute of Technology at Weihai, China
Mallipeddi Rammohan Nanyang Technological University, Singapore
Yazhou Ren University of Science and Technology of China, China
Ko Sakai University of Tsukuba, Japan
Shunji Satoh The University of Electro-Communications, Japan
Gerald Schaefer Loughborough University, UK
Sachin Sen Unitec Institute of Technology, New Zealand
Hamid Sharifzadeh Unitec Institute of Technology, New Zealand
Nabin Sharma University of Technology Sydney, Australia
Yin Sheng Huazhong University of Science and Technology,
China
Jin Shi Nanjing University, China
Yuhui Shi Southern University of Science and Technology, China
Hayaru Shouno The University of Electro-Communications, Japan
Ferdous Sohel Murdoch University, Australia
Jungsuk Song Korea Institute of Science and Technology
Information, South Korea
Andreas Stafylopatis National Technical University of Athens, Greece
Jérémie Sublime ISEP, France
Ponnuthurai Suganthan Nanyang Technological University, Singapore
Fuchun Sun Tsinghua University, China
Ning Sun Nankai University, China
XIV ICONIP 2018 Organization
Norikazu Takahashi Okayama University, Japan

Ken Takiyama Tokyo University of Agriculture and Technology,
Japan
Tomoya Tamei Kobe University, Japan
Hakaru Tamukoh Kyushu Institute of Technology, Japan
Choo Jun Tan Wawasan Open University, Malaysia
Shing Chiang Tan Multimedia University, Malaysia
Ying Tan Peking University, China
Gouhei Tanaka The University of Tokyo, Japan
Ke Tang Southern University of Science and Technology, China
Xiao-Yu Tang Zhejiang University, China
Yang Tang East China University of Science and Technology,
China
Qing Tao Chinese Academy of Sciences, China
Katsumi Tateno Kyushu Institute of Technology, Japan
Keiji Tatsumi Osaka University, Japan
Kai Meng Tay Universiti Malaysia Sarawak, Malaysia
Chee Siong Teh Universiti Malaysia Sarawak, Malaysia
Andrew Teoh Yonsei University, South Korea
Arit Thammano King Mongkut’s Institute of Technology Ladkrabang,
Thailand
Christos Tjortjis International Hellenic University, Greece
Shibata Tomohiro Kyushu Institute of Technology, Japan
Seiki Ubukata Osaka Prefecture University, Japan
Eiji Uchino Yamaguchi University, Japan
Wataru Uemura Ryukoku University, Japan
Michel Verleysen Universite catholique de Louvain, Belgium
Brijesh Verma Central Queensland University, Australia
Hiroaki Wagatsuma Kyushu Institute of Technology, Japan
Nobuhiko Wagatsuma Tokyo Denki University, Japan
Feng Wan University of Macau, SAR China
Bin Wang University of Jinan, China
Dianhui Wang La Trobe University, Australia
Jing Wang Beijing University of Chemical Technology, China
Jun-Wei Wang University of Science and Technology Beijing, China
Junmin Wang Beijing Institute of Technology, China
Lei Wang Beihang University, China
Lidan Wang Southwest University, China
Lipo Wang Nanyang Technological University, Singapore
Qiu-Feng Wang Xi’an Jiaotong-Liverpool University, China
Sheng Wang Henan University, China
Bunthit Watanapa King Mongkut’s University of Technology, Thailand
Saowaluk Watanapa Thammasat University, Thailand
Qinglai Wei Chinese Academy of Sciences, China
Wei Wei Beijing Technology and Business University, China
Yantao Wei Central China Normal University, China
ICONIP 2018 Organization XV
Guanghui Wen Southeast University, China

Zhengqi Wen Chinese Academy of Sciences, China
Hau San Wong City University of Hong Kong, Hong Kong SAR,
China
Kevin Wong Murdoch University, Australia
P. K. Wong University of Macau, SAR China
Kuntpong Woraratpanya King Mongkut’s Institute of Technology
Chaokuntaharn Ladkrabang, Thailand
Dongrui Wu Huazhong University of Science and Technology,
China
Si Wu Beijing Normal University, China
Si Wu South China University of Technology, China
Zhengguang Wu Zhejiang University, China
Tao Xiang Chongqing University, China
Chao Xu Zhejiang University, China
Zenglin Xu University of Science and Technology of China, China
Zhaowen Xu Zhejiang University, China
Tetsuya Yagi Osaka University, Japan
Toshiyuki Yamane IBM, Japan
Koichiro Yamauchi Chubu University, Japan
Xiaohui Yan Nanjing University of Aeronautics and Astronautics,
China
Zheng Yan University of Technology Sydney, Australia
Jinfu Yang Beijing University of Technology, China
Jun Yang Southeast University, China
Minghao Yang Chinese Academy of Sciences, China
Qinmin Yang Zhejiang University, China
Shaofu Yang Southeast University, China
Xiong Yang Tianjin University, China
Yang Yang Nanjing University of Posts and Telecommunications,
China
Yin Yang Hamad Bin Khalifa University, Qatar
Yiyu Yao University of Regina, Canada
Jianqiang Yi Chinese Academy of Sciences, China
Chengpu Yu Beijing Institute of Technology, China
Wen Yu CINVESTAV, Mexico
Wenwu Yu Southeast University, China
Zhaoyuan Yu Nanjing Normal University, China
Xiaodong Yue Shanghai University, China
Dan Zhang Zhejiang University, China
Jie Zhang Newcastle University, UK
Liqing Zhang Shanghai Jiao Tong University, China
Nian Zhang University of the District of Columbia, USA
Tengfei Zhang Nanjing University of Posts and Telecommunications,
China
Tianzhu Zhang Chinese Academy of Sciences, China
XVI ICONIP 2018 Organization
Ying Zhang Shandong University, China

Zhao Zhang Soochow University, China
Zhaoxiang Zhang Chinese Academy of Sciences, China
Dongbin Zhao Chinese Academy of Sciences, China
Qiangfu Zhao University of Aizu, Japan
Zhijia Zhao Guangzhou University, China
Jinghui Zhong South China University of Technology, China
Qi Zhou University of Portsmouth, UK
Xiaojun Zhou Central South University, China
Yingjiang Zhou Nanjing University of Posts and Telecommunications,
China
Haijiang Zhu Beijing University of Chemical Technology, China
Hu Zhu Nanjing University of Posts and Telecommunications,
China
Lei Zhu Unitec Institute of Technology, New Zealand
Pengefei Zhu Tianjin University, China
Yue Zhu Nanjing University, China
Zongyu Zuo Beihang University, China
Contents – Part II
Other Neural Network Models
Improved Kernel Density Estimation Self-organizing Incremental Neural

Network to Perform Big Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Wonjik Kim and Osamu Hasegawa
HISBmodel: A Rumor Diffusion Model Based on Human Individual

and Social Behaviors in Online Social Networks . . . . . . . . . . . . . . . . . . . . . 14
Adil Imad Eddine Hosni, Kan Li, and Sadique Ahmed
Multi-scale Feature Decode and Fuse Model with CRF Layer

for Boundary Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Zihao Dong, Ruixun Zhang, Xiuli Shao, Huichao Li, and Zihan Yang
Sentimental Analysis for AIML-Based E-Health Conversational Agents . . . . . 41

David Ireland, Hamed Hassanzadeh, and Son N. Tran
Hashtag Recommendation with Attention-Based Neural Image

Hashtagging Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Gaosheng Wu, Yuhua Li, Wenjin Yan, Ruixuan Li, Xiwu Gu,
and Qi Yang
CocoNet: A Deep Neural Network for Mapping Pixel Coordinates

to Color Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Paul Andrei Bricman and Radu Tudor Ionescu
BCMLP: Binary-Connected Multilayer Perceptrons . . . . . . . . . . . . . . . . . . . 77

Ningqi Luo, Binheng Song, Yinxu Pan, and Bin Shen
Network of Recurrent Neural Networks: Design for Emergence . . . . . . . . . . 89

Chaoming Wang and Yi Zeng
Computationally Efficient Radial Basis Function . . . . . . . . . . . . . . . . . . . . . 103

Adedamola Wuraola and Nitish Patel
A Model for Age and Gender Profiling of Social Media Accounts Based
on Post Contents. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Jan Kristoffer Cheng, Avril Fernandez, Rissa Grace Marie Quindoza,
Shayane Tan, and Charibeth Cheng
A Wiener Causality Defined by Relative Entropy . . . . . . . . . . . . . . . . . . . . 124

Junya Chen, Jianfeng Feng, and Wenlian Lu
XVIII Contents – Part II
Scene Graph Generation Based on Node-Relation Context Module . . . . . . . . 134

Xin Lin, Yonggang Li, Chunping Liu, Yi Ji, and Jianyu Yang
Cross-Layer Convolutional Siamese Network for Visual Tracking . . . . . . . . . 146

Yanyin Chen, Xing Chen, Huibin Tan, Xiang Zhang, Long Lan,
Xuhui Huang, and Zhigang Luo
A Data Augmentation Model Based on Variational Approach. . . . . . . . . . . . 157

Lei Xia, Jiancheng Lv, and Yong Xu
Asynchronous Value Iteration Network . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

Zhiyuan Pan, Zongzhang Zhang, and Zixuan Chen
FVR-SGD: A New Flexible Variance-Reduction Method for SGD

on Large-Scale Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Mingxing Tang, Zhen Huang, Linbo Qiao, Shuyang Du, Yuxing Peng,
and Changjian Wang
A Neural Network Model for Gating Task-Relevant Information

by Rhythmic Oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
Ryo Tani and Yoshiki Kashimori
A Hybrid Model Based on the Rating Bias and Textual Bias

for Recommender Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Jiao Dai, Mingming Li, Songlin Hu, and Jizhong Han
Phase and Amplitude Modulation in a Neural Oscillatory Model

of the Orientation Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
Bhadra S. Kumar, Avinash Kori, Sundari Elango,
and V. Srinivasa Chakravarthy
Neural Networks Models for Analyzing Magic: The Gathering Cards . . . . . . 227
Felipe Zilio, Marcelo Prates, and Luis Lamb
MulAttenRec: A Multi-level Attention-Based Model for Recommendation . . . 240

Zhipeng Lin, Wenjing Yang, Yongjun Zhang, Haotian Wang,
and Yuhua Tang
SSteGAN: Self-learning Steganography Based on Generative

Adversarial Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
Zihan Wang, Neng Gao, Xin Wang, Xuexin Qu, and Linghui Li
A Multidimensional Interaction-Focused Model for Ad-Hoc Retrieval . . . . . . 265

Qiang Sun, JiaLiang Wu, and Yue Wu
Accounting Results Modelling with Neural Networks: The Case

of an International Oil and Gas Company . . . . . . . . . . . . . . . . . . . . . . . . . 275
Yang Duan, Chung-Hsing Yeh, and David L. Dowe
Contents – Part II XIX
Attention Based Dialogue Context Selection Model. . . . . . . . . . . . . . . . . . . 286

Weidi Xu, Yong Ren, and Ying Tan
Memory-Based Model with Multiple Attentions for Multi-turn

Response Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
Xingwu Lu, Man Lan, and Yuanbin Wu
A Robust LPNN Technique for Target Localization Under Hybrid

TOA/AOA Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
Muideen Adegoke, Andrew Chi Sing Leung, and John Sum
Central Pattern Generator Based on Interstitial Cell Models Made

from Bursting Neuron Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
Takahiro Toizumi and Katsutoshi Saeki
Active Feedback Framework with Scan-Path Clustering for Deep

Affective Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330
Li-Ming Zhao, Xin-Wei Li, Wei-Long Zheng, and Bao-Liang Lu
Stability Analysis
A New Robust Stability Result for Delayed Neural Networks . . . . . . . . . . . . 343

Ozlem Faydasicok, Cemal Cicek, and Sabri Arik
A Novel Criterion for Global Asymptotic Stability of Neutral-Type Neural

Networks with Discrete Time Delays. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
Ozlem Faydasicok and Sabri Arik
Anti-synchronization of Neural Networks with Mixed Delays . . . . . . . . . . . . 361

Dan Liu and Dan Ye
Incremental Stability of Neural Networks with Switched Parameters

and Time Delays via Contraction Theory of Multiple Norms . . . . . . . . . . . . 371
Hao Qiang and Wenlian Lu
Lag Synchronization of Complex-Valued Neural Networks

with Time Delays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
Jiarong Li, Haijun Jiang, Cheng Hu, and Juan Yu
Continuous Attractors of Nonlinear Neural Networks with Asymmetric

Connection Weights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
Jiali Yu, Zhang Yi, Chunxiao Wang, Yong Liao, and Zhixin Pang
XX Contents – Part II
Optimization
Two Matrix-Type Projection Neural Networks for Solving Matrix-Valued

Optimization Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
Lingmei Huang, Youshen Xia, and Songchuan Zhang
An Adaptive Ant Colony System for Public Bicycle Scheduling Problem. . . . 417
Di Liang, Zhi-Hui Zhan, and Jun Zhang
An Artificial Neural Network for Distributed Constrained Optimization . . . . . 430

Na Liu, Wenwen Jia, Sitian Qin, and Guocheng Li
An Estimation of Distribution Algorithm for Large-Scale Optimization

with Cooperative Co-evolution and Local Search . . . . . . . . . . . . . . . . . . . . 442
Jia-Ying Lin, Wei-Neng Chen, and Jun Zhang
A Collaborative Neurodynamic Approach to Symmetric Nonnegative

Matrix Factorization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453
Hangjun Che and Jun Wang
Modularity Maximization for Community Detection Using

Genetic Algorithm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463
Hu Lu and Qi Yao
Continuous Trade-off Optimization Between Fast and Accurate Deep

Face Detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473
Petru Soviany and Radu Tudor Ionescu
Multi-Dimensional Optical Flow Embedded Genetic Programming

for Anomaly Detection in Crowded Scenes . . . . . . . . . . . . . . . . . . . . . . . . 486
Zeyu Mi, Lin Shang, and Bing Xue
Robust Regression with Nonconvex Schatten p-Norm Minimization . . . . . . . 498

Deyu Zeng, Ming Yin, Shengli Xie, and Zongze Wu
Adaptive Crossover Memetic Differential Harmony Search for Optimizing

Document Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509
Ibraheem Al-Jadir, Kok Wai Wong, Chun Che Fung, and Hong Xie
Neurodynamics-Based Nonnegative Matrix Factorization for Classification. . . 519

Nian Zhang and Keenan Leatham
A Multi-kernel Semi-supervised Metric Learning Using Multi-objective

Optimization Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 530
Rakesh Kumar Sanodiya, Sriparna Saha, and Jimson Mathew
Software-as-a-Service Composition in Cloud Computing Using

Genetic Algorithm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 542
Samuel Yu Toh and Maolin Tang
Contents – Part II XXI
Evolving Computationally Efficient Hashing for Similarity Search . . . . . . . . 552

David Iclanzan, Sándor Miklós Szilágyi, and László Szilágyi
Combining Two-Phase Local Search with Multi-objective Ant

Colony Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564
Chun-Wa Leung, Sin-Chun Ng, and Andrew K. Lui
A Neural Network Based Global Optimal Algorithm for Unconstrained

Binary Quadratic Programming Problem . . . . . . . . . . . . . . . . . . . . . . . . . . 577
Shenshen Gu and Xinyi Chen
Supervised Learning
Stochasticity-Assisted Training in Artificial Neural Network . . . . . . . . . . . . . 591

Adedamola Wuraola and Nitish Patel
An Independent Approach to Training Classifiers on Physiological Data:

An Example Using Smiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 603
Md Zakir Hossain and Tom D. Gedeon
Mixed Precision Weight Networks: Training Neural Networks with Varied

Precision Weights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 614
Ninnart Fuengfusin and Hakaru Tamukoh
Policy Space Noise in Deep Deterministic Policy Gradient . . . . . . . . . . . . . . 624

Yan Yan and Quan Liu
Mixup of Feature Maps in a Hidden Layer for Training of Convolutional

Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 635
Hideki Oki and Takio Kurita
Deep Deterministic Policy Gradient with Clustered Prioritized Sampling . . . . 645

Wen Wu, Fei Zhu, YuChen Fu, and Quan Liu
Induced Exploration on Policy Gradients by Increasing Actor Entropy

Using Advantage Target Regions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 655
Alfonso B. Labao, Carlo R. Raquel, and Prospero C. Naval Jr.
MCP Based Noise Resistant Algorithm for Training RBF Networks

and Selecting Centers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 668
Hao Wang, Andrew Chi Sing Leung, and John Sum
Fault-Resistant Algorithms for Single Layer Neural Networks. . . . . . . . . . . . 680

Muideen Adegoke, Andrew Chi Sing Leung, and John Sum
Adversarial Minimax Training for Robustness Against

Adversarial Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 690
Ryota Komiyama and Motonobu Hattori
XXII Contents – Part II
SATB-Nets: Training Deep Neural Networks with Segmented Asymmetric

Ternary and Binary Weights. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 700
Shuai Gao, JunMin Wu, Da Chen, and Jie Ding
Asynchronous Methods for Multi-agent Deep Deterministic

Policy Gradient. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 711
Xuesong Jiang, Zhipeng Li, and Xiumei Wei
A Revisit of Reducing Hidden Nodes in a Radial Basis Function Neural

Network with Histogram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 722
Pey Yun Goh, Shing Chiang Tan, and Wooi Ping Cheah
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 733

Other Neural Network Models
Improved Kernel Density Estimation
Self-organizing Incremental Neural Network
to Perform Big Data Analysis
Wonjik Kim1(&) and Osamu Hasegawa1,2

1
Department of Systems and Control Engineering, Tokyo Institute of
Technology, Tokyo, Japan
{kim.w.ab,hasegawa.o.aa}@m.titech.ac.jp
2
Inc.SOINN, Cureindo-Building 405, Turuma8-4-30, Tamachi, Tokyo, Japan
[email protected]
Abstract. Plenty of data are generated continuously due to the progress in the
field of network technology. Additionally, some data contain substantial noise,
while other data vary their properties in according to various real time scenarios.
Owing to these factors, analyzing big data is difficult. To address these prob-
lems, an adaptive kernel density estimation self-organizing neural network
(AKDESOINN) has been proposed. This approach is based on the kernel
density estimation self-organizing incremental neural network (KDESOINN),
which is an extension of the self-organizing incremental neural network
(SOINN). An SOINN can study the distribution using the input data online,
while KDESOINN can estimate the probability density function based on this
information. The AKDESOINN can adapt itself to the changing data properties
by estimating the probability density function. Further, the experimental results
depict that AKDESOINN succeeds in maintaining the performance of KDE-
SOINN, while depicting an ability to adapt to the changing data.
Keywords: Neural network Kernel density estimation Data analysis

Self-organizing incremental neural network
1 Introduction
Due to the expansion of network communications, data are generated continuously.

Such data are called big data, and there have been many attempts to analyze and apply
them to various research fields [1, 2, 11, 16, 24].
Laney derived three concepts of the big data characteristics, which are as follows [18]:
(1) Data Volume: Massive amounts of data that continue to grow after being
generated.
(2) Data Velocity: Increasing numbers of networks are generating data continuously,
which means that the data generation velocity is very high.
(3) Data Variety: Data in a pool can be of different types, such as time-series, real
environment, artificial environment, textual, and image data.
© Springer Nature Switzerland AG 2018

L. Cheng et al. (Eds.): ICONIP 2018, LNCS 11302, pp. 3–13, 2018.
https://doi.org/10.1007/978-3-030-04179-3_1
4 W. Kim and O. Hasegawa
These three characteristics that are considered by [18] are known as the 3Vs, and
are taken into consideration while dealing with big data.
In machine learning and data analysis research, it is necessary to estimate the
probability density. However, it is difficult to estimate the probability density of big
data due to the three reasons [22].
First, the density estimator for big data must be nonparametric because of the data
volume. Further, we observe that parametric methods are effective for handling fixed
data, because it is possible to tune the parameters of the method to obtain an optimal
performance. However, the volume of big data is not observed to be constant.
Therefore, the volume of big data cannot be analyzed in advance in order to obtain
optimal parameters for the density estimator. However, we observe that the nonpara-
metric density estimator is not troublesome, since analyzing and constructing a big data
model beforehand is not necessary for a nonparametric density estimator.
Second, the density estimator for big data must use online learning methods due to
the observed data velocity. In big data, massive amounts of data grow quickly until the
total size of data becomes gigantic. Online learning methods can be sequentially
updated using the growing data.
Third, the density estimator for big data must be robust. Data that are collected from
real environments often contain noise, which could cause overfitting and decrease
performance. Thus, robust methods are required to deal with data that contain noise.
Further, we observe that robustness is defined differently across various fields [9,
13]. In this study, we define robustness as ‘a function that provides almost the same
results as learning data without noise when learning with noisy data.’ [22]. Further, we
observe that there are two types of noise. The first type is the noise that is generated by
the environment, but that is not related to the objective distribution. Thus, this type of
noise needs to be eliminated. The second type is observed to be related to variance and
fluctuation. Therefore, this type of noise must be preserved.
The kernel density estimation self-organizing incremental neural network (KDE-
SOINN) method [22] satisfies all the three conditions for dealing with big data and is
further observed to be robust to noise. However, it cannot adapt to a changing envi-
ronment. Due to the variety of big data, the structure of data is likely to vary at any
instance. Therefore, an ability to adapt to the observed variation of data is required. In
this study, we propose a revised KDESOINN method to solve this problem. Further,
our proposed method has been termed adaptive KDESOINN (AKDESOINN) in this
paper.
2 Related Works
2.1 Kernel Density Estimation

Kernel Density Estimation (KDE) is a typical nonparametric density estimation
approach [23]. The methodology of KDE process is presented in Algorithm 1
Improved Kernel Density Estimation Self-organizing Incremental Neural Network 5
Algorithm 1. Kernel Density Estimation

(1) Require: training samples xi xi 2 Rd ; i ¼ 1; 2; . . .; N , K : kernel function,
H : bandwidth matrix
P
(2) ^pð xÞ ¼ N1 Ni¼1 KH ðx xi Þ
For the kernel function K, the Gaussian kernel is often used in an identical manner
as that in (1)
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

KH ðx lÞ ¼ 1= ð2pÞd jH j exp ðx lÞT H 1 ðx lÞ=2 ð1Þ
H in algorithm 1 is a parameter, which influences the performance of the estimation

function. Further, attempts have been made to optimize the estimation function [8, 10].
KDE has been investigated using several methods such as by method of setting the
number of kernels [3], gradient descent method [19], and online clustering method [17].
2.2 Self-organizing Incremental Neural Network

In the field of artificial intelligence, artificial neural networks have been recently
proposed. They are usually classified into two groups, namely, supervised and unsu-
pervised learning [25].
SOINN is an unsupervised learning method that is driven by growing neural gas
[4]. There are several kinds of SOINN, including two-layer [5], enhanced [6], and
adjusted SOINN [7]. Since the adjusted SOINN has less parameters than that of the
other SOINNs, it is generally used in applied research [12, 14, 15].
While SOINN learns from the training data, it constructs a data network through
competitive learning. Various nodes are added or deleted from the network or they may
update their location. Further, the edges are added or deleted in a similar manner as the
nodes. Thus, the SOINN network is updated in order to approximate the distribution
using the added input data.
The flowchart of the adjusted SOINN is depicted in Fig. 1, and its procedural flow
is presented in Algorithm 2
Algorithm 2. Adjusted SOINN process
(1) Require: A: set of all neurons. C A A: set of all edges. Ni : set of all
neighbors of neuron i. Wi : weight of neuron i. k: time period to delete redundant
neurons. agemax : parameter to delete edges.
(2) if first time of input then
(3) A c1 ; c2 ; randomly pick up two vectors from training data to initialize the
neuron set.
(4) C ;
(5) end if
(6) while input data n exist do
(7) s1 argminc2A kn Wc k: find out the winner.
(8) s2 argminc2Ans1 kn Wc k: find out second winner.
6 W. Kim and O. Hasegawa
Fig. 1. Flowchart of SOINN
(9) calculate similarity thresholds Ts1 ; Ts2 . If i got neighbors, Ti is the distance to the
farthest neighbor, else the distance to the nearest neuron.
(10) if kn Ws1 k [ Ts1 or kn Ws2 k [ Ts2 then
(11) A A [ n: insert n as a new neuron.
(12) else
(13) if ðs1 ; s2 Þ 62 C: there is no edge between the winner and second winner, then
(14) C C [ ðs1 ; s2 Þ: add new edge into the network
(15) end if
(16) ageðs1 ;s2 Þ 0: reset the age of ðs1 ; s2 Þ
(17) ageðs1 ;iÞ ageðs1 ;iÞ þ 1ð8i 2 Nsi Þ: increase age of edges connected with the
winner by 1.
(18) DWsi ¼ ðts1 Þðn Ws1 Þ; DWi ¼ ð100ti Þðn Wi Þð8i 2 Nsi Þ; ðtÞ ¼ 1t
(19) using vartriangleWsi ; DWi to adjust the winner and its neighbors
(20) delete edges whose age is larger than agemax
(21) among these neurons which the edge deleted in last step connected to, delete
neurons having no neighbors.
(22) end if
(23) if input data number becomes n kðn 2 N þ Þ then
(24) Delete neurons having less than one neighbor
(25) end if
(26) end while
Improved Kernel Density Estimation Self-organizing Incremental Neural Network 7
2.3 KDESOINN
KDESOINN is an extended version of the adjusted SOINN [22]. It determines the
structure of the network using each kernel in the node of a local network that is located
near the node. Additionally, it estimates the probability function using the sum of the
kernels. In the adjusted SOINN, only the Euclidean distance is used for calculating the
similarity thresholds. Conversely, KDESOINN calculates the threshold using
Algorithm 3.
Algorithm 3. KDESOINN threshold calculation
(1) Require: A: set of all neurons. n: new sample data. Pi : set of nodes connected to
node i. q: parameter for threshold. -i 2 Rd : positional vector of node i. ti : number
of wins of node i in competitive learning. I: identity matrix. Hi : threshold region
of node i.
minp2Pi wp wi ðPi 6¼ /Þ
(2) calculate ci ¼
minp2Afig wp wi ðotherwiseÞ
P
(3) TPi i2P t i
Pi T
p2Pi tp wp wi wp wi
1
(4) Ci TP i
(5) Mi Ci þ qci I
(6) threshold region Hi ¼ ðn wi ÞT Mi1 ðn wi Þ 1
KDESOINN can divide clusters more effectively than the adjusted SOINN. The
entire process of KDESOINN is presented in Algorithm 4
Algorithm 4. KDESOINN process
(1) Require: A: set of all neurons. C A A: set of all edges. Ni : set of all
neighbors of neuron i. Wi : weight of neuron i. k: time period to delete redundant
neurons. agemax : parameter to delete edges. Pi : set of nodes connected to node i.
q: parameter for threshold. ti : number of wins of node i in competitive learning.
I: identity matrix. E(G): set of edges in graph G.
(2) if first time of input then
(3) A c1 ; c2 ; randomly pick up two vectors from training data to initialize the
neuron set.
(4) C ;
(5) end if
(6) while input data n exist do
(7) s1 argminc2A kn Wc k: find out the winner.
(8) s2 argminc2Ans1 kn Wc k: find out second winner.
(9) calculate similarity thresholds Hs1 ; Hs2 by algorithm 3.
(10) if ðn Ws1 ÞT Ms1 1
ðn Ws2 Þ [ 1 or ðn Ws2 ÞT Ms1
2
ðn Ws2 Þ [ 1 then
(11) A A [ n: insert n as a new neuron.
(12) else
(13) if ðs1 ; s2 Þ 62 C: there is no edge between the winner and second winner, then
(14) C C [ ðs1 ; s2 Þ: add new edge into the network
(15) end if
Exploring the Variety of Random
Documents with Different Content
astonished by the manner in which he had met Fulsom; yet he was
not astonished that the sheriff had appeared. Fiction to the contrary,
every abnormal detail of life in civilized communities involves a
consequence; for what we call civilization is simply the ways of men
set in a groove, and any departure from that groove brings
investigation.
With this intangible flash of mind to mind, with this singular “feel”
that something unsaid lay behind that question, Hardrock considered
briefly and then answered it in utmost frankness.
“Sheriff, if I told you all I knew or thought about it, the chances are
that you’d arrest me.”
Fulsom gave him a glance, and grinned.
“I’d have a hell of a job doin’ it, wouldn’t I—not to mention gettin’
you off to jail?”
Hardrock broke into a laugh. “Good for you! Here’s what I know.”
And he told what had happened to him since arriving on Beaver
Island.
Sheriff Fulsom listened to the story without a word, puffing as

methodically after his pipe had smoked out as before; he sat like an
image of bronze, giving no sign of what was passing in his mind.
With such a man Hardrock was at his ease, for he knew now that he
might expect some measure of justice, and not hasty jumping at
conclusions for the sake of political prestige.
“You got your nerve to tell me all this,” said Fulsom, when he had
finished.
Hardrock knocked out his pipe and filled it anew. “No witnesses
present. Besides, I figure you as square.”
“That’s the hell of it—I got to be square all around. You’re under
arrest for that shootin’, Hardrock Callahan.”
“Eh?” Hardrock stared, for the Sheriff had not moved an inch.
“You’re in earnest?”
“Yep, so far as it goes.” Fulsom wiped his mustache and chuckled.
“Got to do it. I been nosing around the hospital, and heard that
wounded man talkin’ in his fever. Mentioned your name. Now, I’m
right well acquainted with the Beavers—too durned well acquainted
to come over here on business without a posse, unless I come
alone. These lads over here may have their faults, but they’re men
clear through. If I come over alone, I get a square deal. If I come with
a posse, I’m liable to get most anything. Well, now, I come over to
look you up and see what I could learn. And, from hearin’ your story,
looks like it’s my duty to arrest you. Any law officer would have to do
it on the evidence.”
“All right,” said Hardrock whimsically. “Then what? You can’t prove
my story.”
“Nope. All I figure on is doin’ my duty and breakin’ square with all
concerned. Now, you’re arrested, and charged with murder. You’re in
my custody. You and me understand each other, I guess. I don’t
believe for a minute that things aint exactly as you’ve told ’em to me,
and I figure to stay right here a spell and help you work ’em out. Let’s
see that there fish-flag.”
Hardrock dived into the tent and looked up the bit of canvas. In his
heart he felt a queer sense of relief, a dropping away of all
oppression. This officer was not to be feared. He was under arrest,
and if nothing turned up, he would have to stand trial, and the
evidence was bound to be bad—yet Fulsom was square, and this
counted for everything.
“I’m mighty glad we met up,” he said as he came back to the fire.
“And I reckon we do understand each other, Sheriff. Here’s the flag.
Know it?”
The Sheriff gave it a glance, then laid it down.
“Yep. Belongs to Johnson Brothers of Ludington. But they aint
fished up around these parts—aint fished at all since last year. Sold
out, lock stock an’ barrel, to some fellows from Escanaba, I heard,
who were carrying on the business. Now, either those fellows are
running nets up this way, which I don’t hardly think is so, or else it’s
like you say—they’re running something else for bigger money.
S’pose you and me go out early in your canoe and look for that fish-
trap. Eh?”
“You’re on,” said Hardrock cheerfully.
CHAPTER VII
The boats went out Monday morning, went out early. They went
out from the St. James harbor and from the scattered holdings on
the other islands, boats of Indians and white men, out to the fishing
grounds where lacy gill-nets and hidden trap-nets and long bloater
lines and other legal and illegal methods of obtaining the finny prey
were put into effect. Boats bobbed here and there against the
horizon of island or sea or reef, and engines whirred as the lifters
brought the nets aboard, while trout and whitefish and perch went
tumbling down into the tubs. There was heavy work to be done,
since the fish must be all cleaned and boxed and in to St. James to
make that afternoon’s mailboat.
All that morning Hardrock’s canoe bobbed here and there off the
end of Hog Island, with a drag out from bow and stem, countering
back and forth. It was too shallow hereabout for the big fish, and the
waters looked all deserted, with only a sparkling flash of gulls off the
blue line that marked the north end of Garden to show that a boat
was working there beneath the horizon.
Back and forth they went, and found nothing, though they
searched hard enough for any sign of the black ropes that might
mark a trap. Nothing came near them on the water, excepting a
covey of young ducks that bore down and then wheeled and went
flashing away through the waves in a hurry. With noon, they returned
to camp, where the Sheriff’s launch was drawn safely out of sight
among the bushes down the shore, and lunched leisurely, and then
returned again to the search.
It was nearly three o’clock when at last they found the trap, and
then only by accident, for one of the drags picked up the mooring
line, and Hardrock hauled the canoe along this until the dim mass of
the trap itself was under the canoe. Fulsom came to his assistance,
since it was no light task to haul in the heavy lines without tipping the
canoe, and together they got it to the surface. They could see perch
in it, and big Bullheads from the mud bottom, and one lordly yellow
sunfish, but no whisky.
“Hold on!” exclaimed Fulsom, who knew more about traps than
did Hardrock. “Hold her till I get a grip on that mooringline! Now let
go, and catch hold.”
Now they tugged at the line, and bit by bit worked loose the
anchor down below, and after a time got it on the up-heave.
Hardrock was leaning far over on the line, depending on Sheriff
Fulsom to balance the canoe, and giving his entire attention to the
rope below him. This came heaving up soggily from the depths, and
presently disclosed another line knotted around it and hanging
straight down.
“Thought so!” came the exultant voice of Fulsom. “Haul in on the
short line, now—”
In another moment the end of this came into sight, and showed a
firmly lashed case of liquor. Hardrock glanced up over his shoulder.
“Want it aboard?”
“If we can get it, yes. No telling how many more cases there are,
but we’ll have to leave ’em for the present. We’ll see what this is—
make sure of it. Looks to me like you needn’t worry about that
murder charge any more. Better move lively, too. Looks like a boat is
heading this way from Beaver. Left my binoculars in camp, so I can’t
tell much.”
Hardrock could not pause to look—he got the box in under the
canoe, then came the ticklish matter of swinging it aboard. This was
finally accomplished, though at imminent danger of capsizing the frail
craft; then he straightened up for a look at the approaching boat. It
was still half a mile distant, and bearing up between the islands as
though heading for them.
“Better get in to shore,” said Fulsom. “I aint anxious to be
recognized around here until it’s necessary, the way things are now.
Looks like we got some Canadian Club here, all right—we’ll open her
up and make sure. Set that extry paddle in the trap to mark her
before we go.”
Hardrock nodded and made fast the paddle so that it floated on
the line from which the whisky-case had been cut, then he headed
the canoe for the point and pushed her hard. Whether that boat was
heading for them or not, he meant to take no chances.
In ten minutes he was cutting through the shallows inside the
point and was out of sight of the boat. When they came to camp,
they speedily lifted the canoe ashore and in among the trees. Then
Fulsom, obtaining Hardrock Callahan’s woods hatchet, began to pry
at the lid of the whisky-case.
“Aren’t you tampering with evidence?” said Hardrock, chuckling.
“Who, me? I aint no prohibition officer,” returned the Sheriff dryly.
“No sir, I never voted for no prohibition, but I aim to do my duty. First
thing is to find out if this stuff is whisky or not. Can’t tell by the box,
can’t tell by the label—”
“The only way is to taste it, eh?” laughed Hardrock. “All right, I’m
with you, and will give expert testimony. Go to it! We can’t afford to
make any mistakes; that’s sure.”
The case opened, Fulsom produced a bottle, unhurt by its
immersion, and attacked the cork. When this was out, he handed the
bottle to his nominal prisoner.
“Let’s have your verdict, Hardrock!”
The latter tasted the contents, and grimaced. “It’s the stuff,” he
returned, handing back the bottle. The Sheriff promptly tilted it, and
held it tilted until his breath was gone. Then, gasping, he lowered it,
and replaced the cork.
“Gosh, that’s good!” he observed. “Wisht I could keep the whole
bottle.”
“Go ahead.”
“Nope.” He slid it back into the case. “I could sort of ease my
conscience by havin’ an excuse for one drink to make certain what
the stuff was. And I sure made that drink a good one! But any more’d
be stealin’ evidence, which I don’t aim to do. S’pose you slip out to
the shore and keep an eye on that there boat. Maybe she’s the one
we’re lookin’ for. I’ll lay up out o’ sight till I see who it is.”
Smiling to himself at the odd conceit of the Sheriff, whose regretful
devotion to duty was indubitably sincere, Hardrock left the cover of
the trees and returned to his clearing. He was just in time to see the
launch which they had observed come circling around the point and
head in. To his astonishment, he saw the figure of Nelly Callahan
standing in the bow, while another figure aft was tending the engine.
The girl waved to him eagerly, while her companion, a young
fellow no more than a boy, shut off the engine and let the boat run in
until her nose touched the sand. By the flush of excitement in the
girl’s face, Hardrock guessed that she carried news of some kind.
She jumped ashore, then turned and waved her hand at the boy.
“Hardrock, this is Tom Boyle Gallagher’s boy Micky—Vesty
Gallagher was sending him over to find you, so I came along to bring
the message myself. I knew more about it than Vesty did, anyway,
because I heard Hughie Dunlevy talking to Father last night—”
“All right,” cut in Hardrock. “Wait just a minute, will you? Come
ashore, Micky. Got any gasoline aboard?”
“Ten gallon in the tank still,” said the boy, grinning.
“Know anything about engines?”
“He knows all about ’em,” broke in the girl. “Why?”
“I have a launch down the shore that I’d like to have him look over.
She’s down by that clump of sumach, Micky, drawn up. See if you
can find the trouble, will you? We may have to put her into the
water.”
“Sure,” and Micky started off. Hardrock turned to the girl, smiling.
“Excuse me for the interruption, but I had a bit of news too, and
didn’t want him to overhear. Now come and sit down and tell me
what’s on your mind.”
They sat down together on a fallen log at the edge of the clearing,
and Hardrock got his pipe alight.
“Two things,” said the girl, “or maybe three,” and she laughed.
“First, Hughie and some of his friends are coming over here tonight. I
heard him tell Father they meant to drive you away, and send you
back to Arizona.”
Hardrock, thinking of the Sheriff among the trees, broke into a
hearty laugh.
“Go on,” he said after a minute. “Go on! What next?”
“Isn’t that enough? Vesty got wind of it, and sent Micky off to warn
you. There’s no telling what they’ll do, really—and it’s nothing to
laugh about!”
“It will be, I promise you,” and Hardrock chuckled. “Not for them to
laugh about, though. Don’t mention it to anyone, for he doesn’t want
it known—but Sheriff Fulsom is over there in the trees now. It’s his
launch that is down the shore. I picked him up last night—he was
drifting up the channel, disabled and out of gas. He and I are working
on this business, and we’ve already proved my ideas right by finding
that fish-trap and a case of whisky with it. There are other cases at
the same spot, probably.”
She stared at him, wide-eyed. “Oh, good!” she exclaimed.
“And I don’t forget that I owe the tip to you, either,” he went on.
“Well, what next?”
“Hughie thinks that you did the shooting, but he isn’t sure. He told
Father that a strange launch had been seen around here—a green
boat with a red stripe running around the house. A fishboat. I thought
right away that it might be the one—”
“Good for you, Nelly Callahan! I’ll bet a dollar she’s the one we’re
looking for. Any further news from the chap who went over to the
hospital?”
“He’s still between life and death, they said.”
“Looks bad. Well, what else is on your mind?”
She looked down at the sand, stirred a branch of ground-cedar
with her foot, colored faintly. Then her eyes, direct and searching,
lifted suddenly to meet his gaze.
“Nothing.”
Hardrock frowned. “Something you don’t want to tell me, you
mean?”
“Yes. Please don’t ask.”
For a moment Hardrock looked into the troubled depths of her

eyes, and the answer came to him. He remembered his talk with her
father; he could make a shrewd guess at about what that sort of a
man would do and say to the girl.
“All right, I wont,” he said abruptly. “You remember what we were
talking about when the boat came along and you had to jump in and
go? About Arizona, and you, and Danny’s picture of you. That’s why
I came up here to the Beavers, Nelly. Now let’s not have any
discussion of the question. I don’t want to know what your father
said, or how he may have reported what I said to him. The facts are
that I came here because I had seen your picture, and now that I’ve
met you, I’m going to stay here for a while. I told your father so, and
it’s nothing to be ashamed of. Here’s Micky coming back, so let’s
drop the subject until a better time. I’ll be taking you to the dance
Thursday night, as the boys say. What’s the good word, Micky?”
The grease-smeared lad grinned widely.
“Ye can’t run an engine without a spark, can ye? Sure, she’s all
right—I’ve got some extry batteries here and can fix her up in no
time.”
“But that wont fix the leaky gas tank.” Hardrock looked at the boy’s
boat—an open launch of no great size. “See here, Micky! Could you
run off some gas into that big tin can aboard your boat, and siphon
that into the carburetor, and run my launch into the harbor? If you
can, there’s a ten-dollar bill for you. Leave your boat here and I’ll rent
it until you can get my tank soldered up.”
“You bet!” exclaimed the youth eagerly. “Half an hour and I’ll have
her in shape. You going back with me, Nelly?”
“Yes, and hurry up,” said the girl. “We don’t want to be out all day
and night.”
Between them, Hardrock and Micky got the Sheriff’s launch back
into the water, and the boy fell to work. There was no occasion to
construct a siphoning arrangement, for he discovered that the leak
lay in the piping connections, and stopped it temporarily with some
soap. When he had run five gallons of gasoline into the tank and
turned over the engine, it functioned perfectly.
“Hop in, Nelly!” he sang out. “We’ll get back ’fore dark.”
“Thank you for coming over, dear girl,” said Hardrock, as he gave
Nelly a hand and helped her into the boat. “If I don’t come around
before then, I’ll see you Thursday night. Good-by, and good luck!”
“Good-by,” she answered quietly. Then, as the boat circled out
from shore, he saw her turn a laughing face, and lift her fingers to
her lips, blowing him a kiss. For a moment he stood astounded, then
a laugh broke from him, and a long shout.
“I may not wait until Thursday—after that!” he called, and she
waved her hand in farewell. Then the launch was drawing around for
the point, and passed from sight.
Sheriff Fulsom appeared from the bushes, and he regarded
Hardrock with twinkling eyes.
“Gosh, ye look right happy over something!” he commented dryly.
“Say, this was a good job ye done, too—got us a launch all
shipshape! They’ll recognize my launch over to St. James, but no
matter. Nobody’ll see it until tomorrow anyhow.”
“You heard what she told me?” demanded Hardrock. The Sheriff
nodded.
“Yep. I don’t know that boat, but no matter. She’s our meat, I
reckon, if she’ll only come and pick up that shipment o’ case goods!
But what about them fellows coming over here tonight?” His shrewd
gaze inspected Hardrock gayly. “Looks to me like you and Dunlevy
are bound to fight it out, young fellow!”
Hardrock chuckled. “We should worry about what happens
tonight. I’m your prisoner and if you don’t protect me— Hello! Sheriff,
where are your binoculars? Get ’em!”
“Gone with my launch, durn you! Why? What you lookin’ at?”
Hardrock, who was staring out to the northeast, drew back from
the shore.
“Looks to me like our boat—see her? Green, sure enough; can’t
tell about the red stripe. Get back out of sight, Fulsom. Here—help
run this launch up a little first! Move sharp. They mustn’t suspect
anyone is here. Can you make her out?”
“Yep. That’s her,” affirmed Fulsom confidently. “Go get your
shotgun, Hardrock.”
CHAPTER VIII
The round ball of the sun was hanging low above the purple line
of Garden Island in the west, and the breeze was down until there
was hardly a ripple on the water. From cover of bushes along the
point, Hardrock and Fulsom watched that green fishboat, a red stripe
running broadly around her, spin past the point and round it, and
head for the floating paddle that marked the whisky-cache.
“She’s fast,” said the Sheriff appraisingly. “Built for the work. She
came up from the south, all right, followed the channel through past
Gray’s Reef as though going to the straits, then cut straight west and
headed here. She wasn’t taking any chances by coming up past
Beaver.”
“What’s your program?” demanded Hardrock.
“Get out in that launch, and get quick. You got your shotgun, I’ve
got my pistol. She’ll let us come alongside, and we’ll grab her, that’s
all. No time to waste. You’re my deputy—swear!”
“I swear,” said Hardrock, and laughed. “Making a prisoner into a
deputy—”
“Oh hell, shove along! We got to move fast. I aim to catch her with
the goods.”
They hurried back along the shore and ran out the open launch.
Fulsom gave his automatic pistol to Hardrock, took the shotgun, and
scrambled into the bow.
“You ’tend the engine. We’ll get ’em back here and put ’em
through the third degree separate. Don’t say a word about the
murder. Leave me to handle it.”
“With pleasure.”
The engine spat and coughed and puffed, and presently they
were slipping out past the long point. The green fishboat had halted
at the fish-trap. She was a boat of fair size, housed over except for
foredeck, after-deck, and a narrow strip along the sides. The after
end of this house was wide open. Forward on each side were wide
openings where the lifter brought in nets and fish.
Just now, however, two men were at work forward in the bow,
hauling in better prey than fish. Several cases were piled up, and
they were getting another case aboard. A third man appeared in the
stern, stared at the launch, and called to his companions. All three
turned, watching her.
Hardrock headed as though to bear up past them for Beaver
Island and waved his hand, to which they made no response. The
man from aft had ducked out of sight, reappearing on the foredeck
with the others. As Fulsom was apparently at work on something and
not interested, the whisky-runners evinced no alarm. Then, when he
was opposite their boat and a hundred feet distant, Hardrock shoved
the tiller hard down and swung in toward her.
One of the three waved his arm and shouted:
“Git away! Sheer off! We don’t want no visitors.”
Sheriff Fulsom straightened up, pointed down, and shouted
something indistinguishable. Hardrock held on his course. Again the
leader of the three waved them off, this time with added oaths.
Fulsom grinned.
“Got something to show ye! Look here—look at this!”
The Sheriff leaned forward as though to drag something up to
sight, then came up with the shotgun leveled. The other boat was
now not thirty feet distant.
“Stand quiet and put your hands up! You’re under arrest. Hands
up, durn ye!”
The whisky-runners were caught entirely unawares. This boat,
obviously an island boat, with only two men in her, had been
unsuspected; while to lake-farers any talk of arrest among the
Beavers was in itself ludicrous. There was nothing ludicrous about
Fulsom or the way he handled his shotgun, however, and after one
surprised oath the astonished and dismayed trio put up their hands.
“Run her alongside,” said the Sheriff to Hardrock. “Then go
aboard and disarm ’em. Go through her for guns. You three gents
roost high and quiet, or I’ll blow daylight into ye.”
“What’s this for, anyhow?” demanded the leader. He was a big,
lantern-jawed fellow marked with a scar across his cheek. His two
comrades were swarthy men, whom Hardrock took to be Greeks or
kindred foreigners. Who are you, holdin’ us up this way?”
“Sheriff,” and Fulsom put up one hand to display his star. “All right,
Hardrock.”
As the two craft came into each other, Hardrock jumped aboard
the larger boat and made fast a line. The sight of the officer’s badge
had disconcerted the trio, and they offered only sullen curses as he
swiftly went through them. From two of them he removed heavy
automatics, which he tossed into his own craft. The third man was
unarmed.
Crawling through the forward opening of the deck-house,
Hardrock paused in surprise. There was no lifter in sight, no nets
were aboard, nor fish. Under him was a pile of a dozen whisky-
cases, the white wood all brown and soggy with water, which had
evidently been picked up at some other point in the course of the
afternoon. A quick search sufficed to show that no rifles or other
weapons were in evidence, and he returned to the foredeck.
“Nothing aboard but whisky, Sheriff, and plenty of that,” he called.
“They loaded another cache aboard before coming here.”
“Right thoughtful of ’em,” said Fulsom grimly, and moved back into
the stern, after tossing the captured weapons ahead of him. “You
three birds hop down into the bow, here. Come along, now, and no
talk.”
“Can’t we fix this up, Sheriff?” demanded the leader. “We got
some money—”
“Now I’ll soak you for attempted bribery,” snapped Fulsom. “Git
down!”
Cursing anew, the scar-faced leader got into the bow of the open
launch, and his two comrades followed him. Fulsom looked up at
Hardrock.
“Cast off that anchor in her bows and make sure the line’s fast.
Give her the len’th. Good holdin’ ground here, and she’ll drift in
toward the shore and set pretty. No wind comin’ up tonight, anyhow. I
got two pair o’ handcuffs at camp, and when we get these birds fixed
up and have supper, we can figger what to do next.”
The three “birds” looked decidedly unhappy. The two Greeks
began to talk in their own language, until the Sheriff peremptorily
shut them up. Hardrock, meantime, dumped the big anchor over the
bows of the green fishboat, watched the line run out until it drew taut,
and then climbed back into his own borrowed craft. The sun was just
sinking from sight.
“Back to camp?” he asked, and Fulsom nodded assent.
The engine started up, and the boat circled out for the point, the
Sheriff standing amidships with his shotgun ready. The three
prisoners, crowded on the bow thwart, showed no symptoms of
putting up any fight, however.
“Simplest thing on earth,” said Fulsom calmly, “is to handcuff a
gent with his arms around a sapling. We’ll do that with two of these
birds, and interview the third—give ’em turn and turn about at it. And
we’ll keep ’em at far separated trees. And no supper. Make ’em talk
better, hungry.”
As they were perhaps meant to do, these words reached and
stung the trio. After a rapid-fire exchange of Greek, the leader turned
around.
“This aint legal!” he exclaimed savagely. “You aint got no warrant
—”
“I got a shotgun,” said the Sheriff, a cold glint in his eyes, “and
you’ll taste it if you get gay. So turn around there and set easy. We
aint ready for you to talk yet awhile.”
The boat was around the point and heading in for the shore.
Hardrock, one hand on the tiller, swept her directly in toward the
clearing, threw out the clutch, and after a moment threw it into
reverse. With hardly a jar, the prow of the boat came into the ground
a couple of feet from shore, weighted down as it was by the three
prisoners.
“Now, then,” ordered Fulsom, “you birds hop out and draw her up.
Don’t any of you make a break, or I’ll pepper your hides!”
The big leader, with a growled oath, obeyed the order. There was
no sand at the water’s edge, the beach being composed of small
stones, which farther back ran into sand. The two Greeks likewise
got out. The leader took the prow, each of the Greeks seized the
gunnel, and they drew up the launch until the bow was on the
shingle.
“Now you, Hardrock,” commanded the Sheriff. “Never mind the
guns—I’ll ’tend to ’em. Run over to my pile of stuff and fetch the
handcuffs, will you?”
“Sure.”
Hardrock stepped past the Sheriff and jumped ashore.
At the same instant, the big leader stooped; and the two Greeks
shoved outward on the boat with all their power. Fulsom, caught
unawares by the tremendous lurch of the boat, lost his balance,
dropped the shotgun, and reeled for an instant. The leader hurled a
chunk of rock that struck the staggering man squarely in the side of
the head and sent him down like a shot.
The whole thing passed off swiftly, neatly, with increditable
precision and accuracy. Even as Hardrock whirled about from his
spring, Fulsom was down and the launch was darting out twenty feet
from shore.
Then he found all three men on top of him. One of the Greeks
came first, and went sprawling in the water as Hardrock’s fist met his
face. The second Greek lunged in from one side, a knife in his hand,
and took a kick under the chin that laid him senseless, but the leader
was hurling himself forward and Hardrock could not evade. Caught
in a burly grip, arms locked, both men went down, thrashing. Even
then, had matters been equal, Hardrock would have won out, for with
a twist he came up on top and rammed a fist into the scarred face—
but just then the first Greek swung a stone that laid the man from
Arizona prostrate. Dazed and almost senseless from the blow,
Hardrock keeled over, and before he could recover he was pinned
down under both opponents.
“Tie him up!” growled the leader, and two minutes later Hardrock
was bound hand and foot, while the Greek stooped over his
unconscious comrade and the burly leader stood laughing and
panting. He grinned down at Hardrock.
“So that’s what we think of you and your blasted Sheriff!” he
declared. “We’ll let him float to Mackinac, if he aint dead. By the time
he gits back here, we’ll sure be on our way. Got a good camp here,
aint you? Guess we’ll git us a bite to eat ’fore we bring up our boat
and beat it.”
For a little, however, the man had his hands full. The groaning
Greek, revived by his compatriot, retrieved his knife and flung
himself on the bound captive; the leader interfered, and the trees
resounded to bellowed oaths and orders and imprecations.
Hardrock, helpless to move, watched and listened grimly. At length
the arguments of the leader took effect.
“And ye don’t want to be the same damned fools ye were before,
do ye?” concluded the wrathful leader. “We don’t want to be trailed
for murder! Leave him be. We’ll fix him so’s he can’t hurt us none—
and we wont murder him neither. Ye may think ye can pull a stunt
like that more’n once, and get away with it; but ye can’t. How d’ye
know that there Sheriff didn’t want ye for the other shootin’, hey?”
The sullen Greek acquiesced, put away his knife, and all three
men stamped away up to the camp. Darkness was gathering upon
the waters, but Hardrock no longer stared after the rapidly vanishing
boat that was drifted off along the shore and toward the open lake.
Those words of the leader were dinning in his brain. He knew now
who had shot down those two boys from St. James.
CHAPTER IX
It was perhaps five minutes afterward, while some tins of food

were being opened, that the three whisky-runners realized they had
committed an error. Their leader, whose name appeared to be
Marks, was the one who realized it most keenly. He came down to
the shore, stared off in the gathering darkness at the boat, now a
mere speck in the dusk, and cursed fervently. The shotgun had gone
into the lake, and their pistols had all floated away with poor Fulsom.
Hardrock chuckled.
“You fellows turn me loose,” he offered, “and I’ll tell you where
there’s a boat laid up down the shore.”
Marks turned away. “You’ll tell more’n that ’fore we’re through with
you. Shut up!”
The three gathered again about their food, getting a fire lighted
and in their clumsy ignorance of the woods heaping on fuel until the
yellow flames were leaping high and far. Over such a fire, any
cookery was impossible, and Hardrock chuckled at their profane
efforts to make coffee without getting the pot too hot to be handled.
He, meantime, while apparently motionless and helpless, was in
reality hard at work. He lay, half sitting, against a log between fire
and shore, at the clearing’s edge, arms bound behind him. He had
been tied up with the first thing to hand—bandanna handkerchiefs
produced by the Greeks, and had made the gratifying discovery that
the material was old and would tear easily. Therefore he was tearing
it, against the log at his back, and by the increasing looseness knew
that his wrists were nearly free.
Marks conferred at length with his companions, who were
obviously taking their orders from him, and presently the two Greeks
rose and stamped off into the darkness along the shore, going
toward the point. Marks himself rolled a cigarette and came toward
Hardrock.
“If you’re going to starve me,” said the latter, “you might at least
starve me on a smoke. Look out your friends don’t get lost.”
Marks laughed easily. “I’ll get you some coffee and a smoke,” he
replied, “if you’ll talk. Will you? Or shall I make you?”
“Sure thing,” exclaimed Hardrock. “It’s a bargain. And cut me
loose.”
“Not much,” retorted the other, and went back to the fire, where he
poured out a tin cup of coffee.
Hardrock seized the instant. His arms came free. Swiftly he got a
hand into his pocket—thus far, they had not searched him except for
weapons—and slid out his pocketknife. His arms again in place
behind him, he opened a blade of the knife, and waited. One cut at
his ankles, and he would be free. Without that cut, he dared take no
chances, tempting as the occasion now was.
For Marks now came back to him, held the lukewarm coffee to his
lips as he drank, then gave him the cigarette and held a match to it.
Sitting down and wiping sweat from his face, for it was hot near that
big fire, the burly ruffian rolled himself another cigarette. He was
almost within arm’s reach of Hardrock—yet the latter controlled
himself. Until his feet were free he must attempt nothing.
“Now let’s have it,” said Marks. “I didn’t want them two lard-eaters
to get wise. What was it the Sheriff wanted to give us the third
degree about?”
“About the shooting you fellows pulled off last time you were
here.”
Marks nodded, a frown darkening his scarred features. Evidently
he had anticipated this information.
“Aint it hell how ye can’t make foreigners savvy anything?” he
demanded, to the astonishment of Hardrock. “Them two fellers have
just one notion o’ fighting—to take a gun and kill somebody! I’ll have
to let ’em go. I can’t make ’em savvy that there’s a durned sight more
danger in a murder charge than in running liquor.”
“You mean they’re working for you?”
“Yep. The blamed fools run on them Beaver men the other day,
found ’em lifting the trap out yonder, and riddled ’em—then let ’em
go. That’s a fool Greek everytime. I wasn’t along, dog-gone it! I was
in Escanaba, sick that day, and ye can’t get nothin’ on me. I got to
stand by them fellers, o’ course, and get ’em away safe, but I don’t
like it a mite. This sort o’ killing is bad business.”
Hardrock laughed curtly. “What about the Sheriff?”
“Oh, him! He’s a Sheriff, takin’ chances. Same with you—depity,
aint ye? Yep. He aint killed, though. He’ll drift over in the channel
and’ll get picked up by a barge. We’ll run ye out to Gull Island and
leave ye there with some grub. That’s decent all around. A fight is
one thing, and killin’ is another thing. I been running booze a year
now, and never had a speck o’ trouble before this. Durn them hot-
headed Greeks! They’ve spoiled the best little game this side the
Soo.”
“You’re sure frank about it,” said Hardrock dryly.”
“Why not? I want you should understand it; I aint anxious to be
follered up for a killin’ I didn’t do! Bad enough to have my business
busted up. Now I got to land this cargo and then go somewheres
else. Dog-gone it! I hope they pass them immygration laws an’ do it
quick. A feller can’t make an honest livin’ no more, the way these
durned foreigners are everywhere.”
Hardrock broke out laughing. Marks surveyed him darkly.
“Ye may think it’s funny, but I don’t. It aint the law so much,
neither. It’s these durned islanders! They’re all over the lakes, them
or their relations. If they take the notion it was me responsible for the
killin’, they’ll drive me off the lakes, that’s what.”
The man’s viewpoint was irresistible, and Hardrock laughed the
harder, while Marks sucked at his cigarette and glowered angrily.
Then came the “chug-chug” of a gas engine, and a low call from the
darkness. Slowly the shape of the green fishboat drifted in upon the
shore and then halted as her bows hit the shallows ten feet from the
beach.
“They had to swim to get her, anyhow!” exclaimed Marks. “The
durned fools needed a bath.” He rose and went past Hardrock to the
shore. “Hey, boys! Toss that anchor ashore so’s she wont drift off.
We’ll get away pretty quick, now.”
Hardrock moved his arm, and the little blade of the penknife
flashed in the firelight as he slashed the bonds about his ankles. He
was free, now—but he must let them all get ashore. His only chance,
against the three of them, was to get their boat and leave them here.
It was a time for strategy, rather than for fighting; so, at least, he
thought. He was to discover his mistake very shortly.
The two Greeks came ashore, bearing a line. It appeared that
they had cut loose the anchor rather than haul it in. There ensued a
furious storm of oaths from Marks; the two men became ugly, and for
a moment it looked as though a row were imminent. Then Marks
cooled down, and told them to get some of the supplies from
Hardrock’s tent aboard the boat. All three passed up to the tent,
none of them observing that the captive was no longer bound.
This was the opportunity Hardrock had been praying for, and he
gathered his muscles. Once he could shove out that boat and
scramble aboard her, he had everything in his own hands! He drew
up his feet, saw that the three men were busily engaged with his
supplies, and rose—
While he was in the very act of rising, a voice boomed out among
the trees at the clearing’s edge:
“There’s Callahan and his whole crowd—git ’em all, lads! Take
’em!”
Hardrock was already springing for the water, but a figure
appeared and blocked him. It was the figure of Hughie Dunlevy.
Instantly, Hardrock realized what had happened, and cursed the luck
that had brought the Beaver lads here at this moment. From the
brush was going up a crash of feet and wild yells, Marks was
bellowing, the Greeks were cursing and fighting—beyond a question,
Dunlevy thought that they were part of a gang under the direction of
Hardrock Callahan.
There was no time for any explanations. The man from Arizona
barely had a chance to check his leap for the water, to spring back
and gain balance, when Dunlevy was upon him with a roar of battle-
fury and a whirl of fists.
“Ye will murder poor lads, will ye?” he yelled, and struck.
Hardrock ducked the blow and answered it with a smash to the
wind that stopped Hughie Dunlevy for an instant. Glancing around,
Hardrock was aware of the three whisky-runners by the tent,
furiously engaged with four or five other men. He and Dunlevy were
for the moment alone. Only a glance—then he was driving at his
opponent, hoping still to get out and aboard the boat.
That hope seemed vain. A wild swing caught Hardrock under the
jaw and knocked him ten feet away; Dunlevy was after him instantly,
leaping high in air to come down upon him boots first. He came
down only on the shingle, however; and the man from Arizona,
evading a savage kick, reached his feet and began to fight.
Hughie Dunlevy gasped and grunted as the blows smashed into
him, while before him in the firelight danced that unhurt face with its
blazing eyes and its furious unleashed anger. For all his tremendous
strength, the islander helplessly gave ground, was driven backward,
fists driving into him with relentless accuracy. In vain he tried to
grapple, to kick, to gouge—each attempt failed and only drew upon
him another terrific smash under the heart.
Wanned as he was by white liquor, having great strength in place
of stamina, Dunlevy could not stand up under this battering. Never
once did Hardrock strike for the face, but drove in fists like hammers
that pounded heart and stomach in frightful repetition.
On the other side of the fire, one Greek was thrashing over the
ground with Jimmy Basset pounding him into submission. Connie
Dunlevy was down, trying to quench a knife slash that ran from
shoulder to elbow. The other three island men were battering Marks,
who was badly hurt and groaning as he fought, and the second
Greek whose knife flashed crimson in the firelight. Now Marks gave
way and came crashing down, and the snarling Greek reeled as a
stone smashed into his face.
Hardrock got home to the wind with one direct punch that sent
Hughie Dunlevy two steps backward and brought down his hands—
drove in another that rocked him, and then set himself deliberately
for the finish. His feet shifting perfectly to keep balance, he now put
over a light tap to the mouth, and then laughed.
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade
Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.
Let us accompany you on the journey of exploring knowledge and

personal growth!
textbookfull.com

Download Full Neural Information Processing: 25th International Conference, ICONIP 2018, Siem Reap, Cambodia, December 13–16, 2018, Proceedings, Part II Long Cheng PDF All Chapters

Uploaded by

Copyright:

Available Formats

Download Full Neural Information Processing: 25th International Conference, ICONIP 2018, Siem Reap, Cambodia, December 13–16, 2018, Proceedings, Part II Long Cheng PDF All Chapters

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Download Full Neural Information Processing: 25th International Conference, ICONIP 2018, Siem Reap, Cambodia, December 13–16, 2018, Proceedings, Part II Long Cheng PDF All Chapters

Uploaded by

Copyright:

Available Formats

Download the Full Version of textbook for Fast Typing at textbookfull.

Neural Information Processing: 25th International

Download More textbook Instantly Today - Get Yours Now at textbookfull.com

Neural Information Processing: 25th International

Neural Information Processing: 25th International

Neural Information Processing: 25th International

Neural Information Processing: 25th International

Neural Information Processing 27th International

Computational Science ICCS 2018 18th International

Neural Information Processing 27th International

Neural Information Processing 27th International

Seiichi Ozawa (Eds.)

ISSN 0302-9743 ISSN 1611-3349 (electronic)

Library of Congress Control Number: 2018960916

LNCS Sublibrary: SL1 – Theoretical Computer Science and General Issues

© Springer Nature Switzerland AG 2018

The 25th International Conference on Neural Information Processing (ICONIP 2018),

October 2018 Jun Wang

Special Sessions Chairs

Ke-Cai Cao Nanjing University of Posts and Telecommunications,

Bo Fan Zhejiang University, China

Panfeng Huang Northwestern Polytechnical University, China

Ya Li Institute of Automation, Chinese Academy of Sciences,

Haruhiko Nishimura University of Hyogo, Japan

Norikazu Takahashi Okayama University, Japan

Guanghui Wen Southeast University, China

Ying Zhang Shandong University, China

Other Neural Network Models

Improved Kernel Density Estimation Self-organizing Incremental Neural

HISBmodel: A Rumor Diffusion Model Based on Human Individual

Multi-scale Feature Decode and Fuse Model with CRF Layer

Sentimental Analysis for AIML-Based E-Health Conversational Agents . . . . . 41

Hashtag Recommendation with Attention-Based Neural Image

CocoNet: A Deep Neural Network for Mapping Pixel Coordinates

BCMLP: Binary-Connected Multilayer Perceptrons . . . . . . . . . . . . . . . . . . . 77

Network of Recurrent Neural Networks: Design for Emergence . . . . . . . . . . 89

Computationally Efficient Radial Basis Function . . . . . . . . . . . . . . . . . . . . . 103

A Wiener Causality Defined by Relative Entropy . . . . . . . . . . . . . . . . . . . . 124

Scene Graph Generation Based on Node-Relation Context Module . . . . . . . . 134

Cross-Layer Convolutional Siamese Network for Visual Tracking . . . . . . . . . 146

A Data Augmentation Model Based on Variational Approach. . . . . . . . . . . . 157

Asynchronous Value Iteration Network . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

FVR-SGD: A New Flexible Variance-Reduction Method for SGD

A Neural Network Model for Gating Task-Relevant Information

A Hybrid Model Based on the Rating Bias and Textual Bias

Phase and Amplitude Modulation in a Neural Oscillatory Model

MulAttenRec: A Multi-level Attention-Based Model for Recommendation . . . 240

SSteGAN: Self-learning Steganography Based on Generative

A Multidimensional Interaction-Focused Model for Ad-Hoc Retrieval . . . . . . 265

Accounting Results Modelling with Neural Networks: The Case

Attention Based Dialogue Context Selection Model. . . . . . . . . . . . . . . . . . . 286

Memory-Based Model with Multiple Attentions for Multi-turn

A Robust LPNN Technique for Target Localization Under Hybrid

Central Pattern Generator Based on Interstitial Cell Models Made

Active Feedback Framework with Scan-Path Clustering for Deep

A New Robust Stability Result for Delayed Neural Networks . . . . . . . . . . . . 343

A Novel Criterion for Global Asymptotic Stability of Neutral-Type Neural

Anti-synchronization of Neural Networks with Mixed Delays . . . . . . . . . . . . 361

Incremental Stability of Neural Networks with Switched Parameters

Lag Synchronization of Complex-Valued Neural Networks