examples

Examples

Tutorials

• knet-tutorial: Linear, MLP, CNN, RNN models; gradients and SGD; overfitting and dropout; sequences and minibatching etc.
• julia-tutorial: Julia examples demonstrating arrays, tuples, dictionaries, indexing etc.
• optimizers: Try various optimizers (SGD, Momentum, Adam...) on LeNet.
• rnn-tutorial: RNN tutorial notebook with BPTT, LSTM, S2S.

Benchmarks

• DeepLearningFrameworks: Notebooks comparing CNTK, Caffe2, Chainer, Gluon, Keras, Knet, Lasagne, MXNet, PyTorch, TensorFlow on CNN and RNN examples.
• dynet-benchmark: Four dynamic neural network examples comparing Knet with DyNet and Chainer from dynet-benchmark.

Models

Linear

• synthetic-linreg: Simple linear regression example using artificial data.
• housing-linreg: Linear regression on the Boston Housing dataset.

MLP

• mnist-mlp: Multi-layer perceptron trained on MNIST.
• fashion-mnist: Multi-layer perceptron trained on Fashion-MNIST.

CNN

• lenet: The LeNet model trained on MNIST.
• cifar10-cnn: CNN model for CIFAR-10 with batchnorm.
• resnet: Knet implementation of ResNet 50, 101, and 152 models.
• vgg: Knet implementation of VGG D and E models.

RNN

• rnnlm: Word-level RNN language model trained on the Mikolov-PTB corpus.

Reinforcement Learning

• dynamic programming : Solving a Markov Decision process with Value Iteration and Policy Iteration.
• policy gradient : Demonstration of the REINFORCE and Actor-Critic algorithms on simple environments.
• dqn : Train a DQN on the CartPole environment.

Other

• variational-autoencoder: Train a Variational Autoencoder on MNIST.
• dcgan-mnist: Train a DCGAN (Deep Convolutional Generative Adversarial Network) on MNIST.