Downloads

Code repository

Download the entire code repository as a zip-file.

Cheat Sheets

Large versions of the DL cheat sheets from Appendix J in the book:

sheet 0

Sample Chapter

Contains table of contents, preface, and a sample chapter:

sample chapter

Alternative version of MNIST examples

If you for some reason cannot get the idx2numpy package working on your platform, then you can download alternative versions of the MNIST examples here.

These versions assume that you have downloaded CSV-versions of the MNIST dataset and placed in the ../data directory. You can find CSV-versions of the MNIST dataset in the following two locations:

https://pjreddie.com/projects/mnist-in-csv

https://www.kaggle.com/oddrationale/mnist-in-csv

Alternative versions of NAS examples

If you run into out-of-memory problems when running the TensorFlow versions of the neural architecture search (NAS) examples, then you can try the following modified versions:

c17e4_nas_random_hill_multiprocess.py

c17e4_nas_random_hill_multiprocess.ipynb

c17e5_nas_evolution_multiprocess.py

c17e5_nas_evoluation_multiprocess.ipynb

These versions spawn a separate process for each model evaluation, and thereby work around a potential memory leak problem (also described in the following link).

Alternative versions of autocompletion examples

If you run into a problem with the application hanging when running the TensorFlow versions of the autocompletion examples, then you can try the following modified versions that don’t use recurrent drop-out:

c11e1_autocomplete_no_rdo.py

c11e1_autocomplete_no_rdo.ipynb

c11e2_autocomplete_embedding_no_rdo.py

c11e2_autocomplete_embedding_no_rdo.ipynb

This is a rare problem, only observed on a single platform at this point.