Scribd
Upload
25 views2 pages

Conference

Uploaded by

ghanatibahar92
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
25 views2 pages

Conference

Uploaded by

ghanatibahar92
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

In the first conference, we introduced a hydrological model, crucial for comprehending and managing water

resources in watershed systems. The methodology we propose consists of two phases :


1) Data Compression: This involves minimizing the data volume necessary to encapsulate a specific
dataset. The objective is to economize storage and reduce the time taken for data transmission across
networks. Research indicates that among deep learning tools, the Autoencoder stands out prominently
for data compression [1]. Autoencoders, a subclass of artificial neural networks, are engineered to
master efficient data encoding autonomously. Their modus operandi is to condense data into a concise,
reduced-dimensional latent space and subsequently reconstruct the original input from this
compressed form. The overarching aim is to diminish any disparity between the initial data and its
reformed version. Indeed, post the autoencoder's training, the latent space (often dubbed the
bottleneck or encoded layer) offers a condensed rendition of the input. This succinct representation
accentuates the data's most significant attributes, effectively functioning as a condensed feature set .
For optimizing the autoencoder's training, a gamut of techniques, such as the Adversarial Generator
Network [2] and Regularization Techniques [3], have been put forth. In this project, I want to use
Adversarial Generator Network for training Autoencoder.
Prediction: This section aims to forecast hydrological parameters. Research indicates that recurrent
neural networks, particularly long short-term memory (LSTM), are fundamental deep learning
architectures for such predictions. For example, Dai et al. [4] proposed a novel LSTM-seq2seq-based
model for short-term hydrological forecasting, addressing common assumptions in time series analysis.
Incorporating feature tests for hydrological time series characteristics and utilizing LSTM for both
encoding historical flows and decoding context vectors, the approach leverages an attention
mechanism for enhanced prediction accuracy. As another example, Anshuka et al. [5] introduced a
deep learning framework using long short-term memory models for spatio-temporal forecasting of
hydrological extremes in the South Pacific, leveraging satellite rainfall estimates and sea surface
temperature anomalies. This study's framework, designed to predict the effective drought index (EDI),
integrates three forecasting approaches, underscoring the importance of understanding dominant
features influencing Pacific precipitation.
In our proposed approach, we aim to leverage enhancements made to the LSTM network, specifically
the transductive long short-term memory (TLSTM) [6], for forecasting purposes. TLSTM acknowledges
the significance of samples near the test point for better model refinement.
We utilize the feature compression achieved in the prior phase as input for the TLSTM network.
We will compare our model with state-of-art models using the CAMELS-GB dataset [7]. CAMELS-GB is a valuable
contribution to the growing collection of large-scale hydrological datasets as it provides catchment attributes
and daily hydrological time series data for 671 catchments in Great Britain. The dataset is publicly available and
can be accessed through the Centre for Ecology & Hydrology (CEH) website, which facilitates its use in research
and applications related to water resources management, climate change, and environmental assessments. The
availability of the CAMELS-GB dataset has allowed researchers to perform comparative studies between
catchments and to develop and test hydrological models at different spatial and temporal scales. The dataset
has also facilitated the investigation of the impacts of land use and climate change on water resources, the
assessment of the effectiveness of water management practices, and the development of early warning systems
for extreme hydrological events.
We intend to participate in the 28th International Conference on Engineering of Complex Computer Systems
(ICECCS 2024), scheduled for June 19-21, 2024 [8]. The deadlines for abstract and full paper submissions are
December 8, 2023, and December 15, 2023, respectively. I aim to provide you with the complete paper for this
conference by the end of November for your review and feedback.
[1] J. Zhai, S. Zhang, J. Chen, and Q. He, "Autoencoder and its various variants," in 2018 IEEE international
conference on systems, man, and cybernetics (SMC), 2018, pp. 415-419: IEEE.
[2] A. Creswell, T. White, V. Dumoulin, K. Arulkumaran, B. Sengupta, and A. A. Bharath, "Generative
adversarial networks: An overview," IEEE signal processing magazine, vol. 35, no. 1, pp. 53-65, 2018.
[3] S. M. Kakade, S. Shalev-Shwartz, and A. Tewari, "Regularization techniques for learning with matrices,"
The Journal of Machine Learning Research, vol. 13, no. 1, pp. 1865-1890, 2012.
[4] Z. Dai, M. Zhang, N. Nedjah, D. Xu, and F. Ye, "A Hydrological Data Prediction Model Based on LSTM
with Attention Mechanism," Water, vol. 15, no. 4, p. 670, 2023.
[5] A. Anshuka, R. Chandra, A. J. Buzacott, D. Sanderson, and F. F. van Ogtrop, "Spatio temporal
hydrological extreme forecasting framework using LSTM deep learning model," Stochastic
Environmental Research and Risk Assessment, vol. 36, no. 10, pp. 3467-3485, 2022.
[6] Y. Yu, X. Si, C. Hu, and J. Zhang, "A review of recurrent neural networks: LSTM cells and network
architectures," Neural computation, vol. 31, no. 7, pp. 1235-1270, 2019.
[7] G. Coxon et al., "CAMELS-GB: hydrometeorological time series and landscape attributes for 671
catchments in Great Britain," Earth System Science Data, vol. 12, no. 4, pp. 2459-2483, 2020.
[8] (2024). 28th International Conference on Engineering of Complex Computer Systems (ICECCS 2024).
Available: https://cyprusconferences.org/iceccs2024/

You might also like