Scribd
Upload
37 views2 pages

Assignment 1B

This document summarizes the key challenges of data management for deep learning systems discussed in an article, including: changes in data sources, expensive data testing, lack of metadata, legislative restrictions on data storage, managing categorical data, feeding dirty data to models, and data shifts affecting model performance. It provides examples of these challenges from real-world applications of deep learning in recommender systems, wind power prediction, and other domains. The document was based on interviews with expert data scientists and practitioners in different fields.

Uploaded by

Kevell Umrao
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
37 views2 pages

Assignment 1B

This document summarizes the key challenges of data management for deep learning systems discussed in an article, including: changes in data sources, expensive data testing, lack of metadata, legislative restrictions on data storage, managing categorical data, feeding dirty data to models, and data shifts affecting model performance. It provides examples of these challenges from real-world applications of deep learning in recommender systems, wind power prediction, and other domains. The document was based on interviews with expert data scientists and practitioners in different fields.

Uploaded by

Kevell Umrao
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

From my understanding after reading the article, the main purpose of this insightful written

document was to define deep learning, distinguish it from other machine learning techniques,
provide real-life situations and thoughts, from different real-life domains, of data management
challenges that companies or practitioners faced during the development of deep learning systems,
highlighting the challenges of data collecting that may effect the performance of deep learning
systems and to provide knowledge and research for companies, the AI community and data
scientists who wish to implement deep learning components or design deep learning models for
present and future reference.

This article was well explained as it provided examples and qualitative data of the data management
challenges that practitioners faced while implementing DL components into real-life applications,
not only from one but various real-life domains such as the: recommender system, wind power
prediction, house price prediction, melanoma detection, financial fraud detection and
manufacturing system, as a means for me, as the reader, to identify and understand where exactly
the main data management issues emerged from in these various situations.

The data collection method for these real-life cases was fair and quite relevant with sources coming
in from expert data scientists and practitioners who are knowledgeable and qualified to provide
insight on data management and the implication and challenges of deep learning methods. After
these expert interviews, the audio recordings of the interviews were summarized and cross-checked
various times, highlighting key points and ensuring relevant and accurate results.

Managing data is such a big thing in our current lives as the world is getting more and more
advanced and digitalized, more data is being released every second on the internet or in books.
Many companies and people in the work environment need data to perform tasks efficiently and
data has to always be stored and processed for efficient use, sharing and management of data.

Some of the main data collecting and management challenges I read were changes in data sources,
expensive data testing, lack of metadata and diverse samples, legislative rules that limit data storage
and processing as a means to protect personal information, duplicate data analyzing and managing
categorical data which are not in numerical form, feeding dirty data to DL models, poor quality data
that puts the performance of results at risk and data shifts and feedback loops.

I do believe deep learning is a growing and exciting technique that will be beneficial to society and
companies and this document is evident that, from the six uses of deep learning systems in real-life
application, DL can be implemented in many different areas of society whether that be for medical
research or health benefits, improve customer service and satisfaction or improve manufacturing
quality of products while reducing the effects of it on the environment and expenses for the
company, etc. But from the document itself, deep learning implementation and data management
has various challenges that I believe will be quite expensive and will require immense planning, time
and research to find solutions to these challenges because as discussed data will have various errors
and outliers and is constantly changing which is what deep learning neural networks and systems
will have to adapt to.

My only remark is that I wish for an article discussing data, that more quantitative data would be
added to give us a more analytical understanding and information on these data management
challenges and the growth of data learning methods. However, the qualitative and case-study
approach was sufficient and informative. It gave a thorough understanding of the real-life DL
implementation challenges from real-life situations and provides a great guide for future research.

You might also like