Homework for Deep Learning class of Charles University.
- numpy_entropy: entropy, cross-entropy, KL-divergence
- mnist_layers_activations: hidden layers, activation functions, accuracy
- mnist_training: optimizers, learning rate, exponential decay
- gym_cartpole: neural network modeling
- mnist_dropout: dropout
- uppercase: one_hot vector
- mnist_conv: CNN, pooling
- mnist_competition: CNN modeling
- mnist_batchnorm: batch normalization
- fashion_masks: image segmentation (pixel bit mask generation)
- 3d_recognition: recognition of 3D objects
- nsketch_transfer: transfer learning
- sequence_classification: RNN modeling, RNN, LSTM, GRU cells, exploding gradient, gradient clipping
- sequence_prediction: low-level handling of RNN cells (dimensionality, type), mean squared error
- tagger_we: part-of-speech tagger, word embeddings, bidirectional RNN, resettable metrics
- tagger_cle: part-of-speech tagger, character-level word embeddings, bidirectional character-level RNN, concatenating word-level embeddings and CLEs
- tagger_cnne: part-of-speech tagger, convolutional embeddings, concatenating word-level embeddings and CNNEs
- tagger_sota: part-of-speech-tagger
- lemmatizer_noattn: lemmatization, characters embedding, training time decoder, inference time decoder
- lemmatizer_attn: lemmatization, bidirectional GRU encoder, attention for encoder
- lemmatizer_sota: lemmatization
- vae: simple Variational Autoencoder
- gan: simple Generative Adversarion Network
- dcgan: Deep Convolutional GAN
- nli: Native Language Identification task on NLI Shared Task 2013 data
- tagger_crf: tagger, bidiretional RNN, Conditional Random Fields layer
- monte_carlo: reinforcement learning using Monte Carlo algorithm
- eager-mnist: mnist dataset, CNN using tensorflow.contrib.eager package
- estimator-mnist: mnist dataset, CNN using tensorflow.estimator package