outline

14 topics by CEATEC

- 001 introduction
-- what is machine learning / deep learning ?
- 002 basic of chainer
-- feedforward networks (example mnist linear, relu, backprop)

specific neural networks

- 003 cnn, image
-- (example ImageNet, convolutional layer, maxpool, - alexnet, resnet
- 004 rnn, nlp
-- example LSTM
-- word2vec (example word2vec, EmbedID)

- 005 deep generative models
-- VAE, GAN

component of neural networks

- 006 activation functions
-- (relu, prelu, elu, sigmoid, tanh)
- 007 regularization
-- (dropout, L2norm)
- 008 optimizers
-- (adagrad, adam, RMSprop)
- faster training (BatchNorm)
- loss (sigmoid_cross_entropy, mean_squared_error)

misc

- numpy, cupy
- using pretrained model (http://qiita.com/tabe2314/items/6c0c1b769e12ab1e2614)
- speech recognition (example maybe translate this into chainer https://github.com/pannous/tensorflow-speech-recognition)
- deep generative models (example VAE)

- k-means and k-means++ by cupy