Artem Bolshakov's repositories
BNN_alternate_SOTA
Reimplementation of the Quicknet from larq.zoo, with no batchnorms
BNN_order_statistics
I use normal distribution order statistics to replace batch normalizations in binary neural networks. I show that this avoids some BatchNorm problems, with little cost.
kArmedBandits
I play with the examples from Ch.1 of my RL book. Specifically, I build the 10-arm platform and test techniques on it.
CliqueProblemV2
Same concept, rotating through a different space.
CliqueProblemWithPruning
Quick NP experiment
discretePlanningFamiliarization
A small space where I become more familiar with PDDL
DLvsCAS
A short summary of my thoughts on Lample and Charton's "Deep Learning for Symbolic Mathematicss," and a comparison to Computer Algebra Systems
gamblingRL
Da Gambling problem from the book
GlobalLocal
Small repo to show the difference between global and local variables
GPUutilityScripts
Small repo for utility scripts I find useful when working with GPUs. Focus is on NVIDIA gpus for AI, but future additions might be more general.
gym-cooking
gym-cooking: Code for "Too many cooks: Bayesian inference for coordinating multi-agent collaboration", Winner of CogSci 2020 conference award in computational modelling.
Language-as-an-Abstraction-for-Hierarchical-Deep-Reinforcement-Learning
PyTorch Implementation of "Language as an Abstraction for Hierarchical Deep Reinforcement Learning" paper
monte-carlo-tree-search
Monte carlo tree search in python
MountainCar
MountainCar problem from Sutton and Barto
mqfigs
Mert Question Figure Backups
Policy-Gradient
Will make example work, then play rock-paper-scissors against slow-learning k-armed bandit.
RLContinuousEval
First Continuous RL module; mostly for evaluation.
RLproper
Monte Carlo, Time Difference methods. Classes for agents that will be used hereafter.
spatial-reasoning
Code for the paper "Representation Learning for Grounded Spatial Reasoning"