Zhihan Yang

off-policy-continuous-control. Official PyTorch code for "Recurrent Off-policy Baselines for Memory-based Continuous Control" (DeepRL Workshop, NeurIPS 21)

gender-audio-classification. A speaker gender classifier. MFC feature engineering and a pre-trained ResNet-50. GradCAM interpretation.

cnnvis. A Python library for visualizing convolutional neural networks.

alpha-zero. Minimal AlphaZero in PyTorch, trained on Connect4 on a 6x6 board.

aevb-tutorial. Minimal VAE, Conditional VAE (CVAE), Gaussian Mixture VAE (GMVAE) and Variational RNN (VRNN) in PyTorch, trained on MNIST.

drqn. Exploring whether DRQN + action prior + state-based expert + history-based entropy-reduction expert

pytorch-sac. Soft actor-critic on Pendulum-v0 with PyTorch

super_mario_as_a_string. Replication of Super Mario as a String: Platformer Level Generation Via LSTMs. Written in Python using Keras.

tmixfit. Vectorized Expectation Maximization for Student-t mixture models in PyTorch.

dior. Differentiable reparameterization of matrices with orthogonal columns.

pytorch-ppo. Minimal proximal policy gradient (clipped version) in PyTorch.

mario-playability-test. Run a simple agent through user-defined super-mario chunks and determine the playability proportion.

pcgml-gmmVae-exp. Implementation of GMM+VAE plus various demos.

pytorch-wgan. Minimal WGAN in PyTorch, trained on 1d and image datasets.

microwave-perceptron. Code and PDF for the project.

cnn-experiments. Code for the project. Under Construction.

math-notes. My math notebook

bishop1995_notes. My notes and demos for understanding Neural Networks for Pattern Recognition (Bishop, 1995).

probml-notes. Notes and answers for Machine Learning: A Probabilistic Perspective

dissonance-quantification. Code and PDF for the project.

d4rl_evaluations. Modification of the original d4rl_evaluations repo for experimentation.

presentations. Presentations in my research group.