I almost finished reading this book: Deep learning with Python [1]. It is a great book for a beginner or programmer aimed on working with deep learning. The book provides a step-by-step guide for solving a variety of problems, mainly classification and regression problems.
Now, I would like to read a book focused on best practices and intuition of ideas (i.e., what to do when a DNN does not work as expected or what to look at when debugging a DNN). That is, the kind of knowledge gained from experience. In some of the projects given in [1], the author suggests how to tackle some of the issues that happen when the model does not work as expected (for instance, in one example he explains why using bidirectional RNN does not improve the performance of a basic RNN).
Although, after searching for a few months, I think that this kind of book is not available yet. As far as I know, Karpathy is the only one that shares this guidelines [2]. However, he does not post frequently anymore. Any suggestions?
[1] https://ift.tt/2k7EoU9 [2] https://ift.tt/2XreASX
Comments URL: https://news.ycombinator.com/item?id=21425100
Points: 1
# Comments: 0
from Hacker News: Newest https://ift.tt/36t5gn5
0 comments:
Post a Comment