movzeroone Goto Github PK

attention-visualized

scaled dot-product attention visualization (single head) (Attention Is All You Need arXiv:1706.03762)

autoencoder

MNIST handwritten digits clustering with an autoencoder

catastrophicforgettingmlp

Catastrophic forgetting on MNIST digets with MLP

dqn

DQN with noisy networks, double and dueling extension. (arXiv:1509.06461, arXiv:1511.06581, arXiv:1706.10295, arXiv:1312.5602)

evolutionary_strategies

based upon (Evolution Strategies as a Scalable Alternative to Reinforcement Learning, arXiv:1703.03864)

flash-attention

Fast and memory-efficient exact attention

focal_loss_visualization

Visualization of the focal loss from Focal Loss for Dense Object Detection arXiv:1708.02002

hebbian_meta_learning

Based upon Meta-Learning through Hebbian Plasticity in Random Networks (arXiv:2007.02686)

idea_book

peliminary idea

learning-iterative-reasoning-through-energy-minimization

My implementation of arXiv:2206.15448

logical_extrapolation

End-to-end Algorithm Synthesis with Recurrent Networks: Logical Extrapolation Without Overthinking arXiv:2202.05826

lorenz-system

maze_generator

implementation of Wilson's algorithm for maze generation

mixture_density_network

multimodal distributions with neural networks

nlp_project

normalizing_flows_with_stochastic_interpolants

Based on building normalizing flows with stochastic interpolants. ArXiv:2209.15571

optimizing_through_surrogate_model

optimizing through surrogate model

pg-is-all-you-need

Policy Gradient is all you need! A step-by-step tutorial for well-known PG methods.

pytorch-tvmisc

Totally Versatile Miscellanea for Pytorch

random_fourier_features

based upon (Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains arXiv:2006.10739)

redq_study_project

remember_exploration

autoencoder with adjustable decorrelation of states using random fourier features

rule-augmented-learning

My implementation of Controlling Neural Networks with Rule Representations ArXiv:2106.07804

siamese_nn_visual

MNIST handwritten digits clustered using siamese nn

zero_competition_gan

Optimizing min-max games using duality gap objective. This code is based upon (Generative Minimization Networks: Training GANs Without Competition arXiv:2103.12685)