Google Group
This Google group is the place students can discuss their projects or questions related to this course.Computing resources
- Paperspace provides discount price of GPU machines for the class ($120/month)
- You are also encouraged to buy a GPU card (GTX 1080 or NVidia Titan X) and install it to your desktop.
Book
There is no required book for this class. But we'd recommend the following reading materials beyond the class:- I. Goodfellow, Y. Bengio, A. Courville, Deep Learning, 2016.
- D. Ferrucci et al, This is Watson, IBM Journal of Research and Development, Vol 56, issue 3-4, 2012.
- B. Christopher, Pattern Recognition and Machine Learning , 2006.
GPU machines
You are enouraged to buy a GPU card (GTX 1080 or NVidia Titan X) and install it to your desktop.
Or you may use the GPU instance on the cloud. Paperspace endorses the students in our class a generous discount to use a dedicated GPU in their cloud.
- Price: $120/month or $0.30/hour
- How to use the discount: register an account on Paperspace and apply the promotion code to your purchase (promotion code to be accounced in the class).
Software
Deep learning frameworks- Keras (suggested)
- Theano
- Tensorflow
- Caffe
- Torch
- MxNet
- PaddlePaddle
Speech Recognition
Deep learning for ASR- Kaldi: http://kaldi-asr.org/
- Kaldi's dnn interface
- Keras + Kaldi
- mxnet + Kaldi
- Baidu warp CTC
- CTC on Kaldi
- Eesen
- Tensorflow’s CTC implementation: only works with sparse tensors
- An example of using CTC for OCR (in Keras)
- Seq2seq in Tensorflow
- Seq2Seq with Keras
Audio Synthesis
Reinforce Learning
- David Silver's tutorial on ICML'16
- Pieter Abbeel's tutorial at Montreal 2016 ML summer school
- Nervana's blog series on RL
- DQN in theano
- OpenAI Gym