Notebooks from Colab
- Chapter 2: keras
- Chapter 2: tf
- Chapter 3: mnist
- Chapter 4: keras for initialization
- Chapter 4: minibatch normalization
- Chapter 5: autoencoders for mnist
- Chapter 5: basic convolutions
- Chapter 6: char-level generation with RNN
- Chapter 7: word2vec
- Chapter 7: word2vec (local, im-memory)
- Chapter 7: word2vec (local, im-memory, load)
- Chapter 8: generative-adversarial network (GAN)
- Chapter 8: adversarial auto-encoder (AAE)
Other links:
- Convolutions arithmetic
- Andrej Karpathy about RNN: very impressive examples, funny quotes:
RNNs are neural networks and everything works monotonically better (if done right) if you put on your deep learning hat and start stacking models up like pancakes.
In case you were wondering, the yahoo url above doesn’t actually exist, the model just hallucinated it.
- vc.ru about app for shot detection. Nice technique for augmentation: add podcasts and bloggers to initial audio track