Home

AgentNet

A lightweight library to build and train neural networks for reinforcement learning using Theano+Lasagne

Installation

An installation guide can be found here

In short,

• Manual install
  • Install bleeding edge theano/lasagne
  • [sudo] pip install --upgrade https://github.com/yandexdataschool/AgentNet/archive/master.zip
• Here's the docker container
  • [sudo] docker run -d -p 1234:8888 justheuristic/agentnet
  • Access via localhost:1234 or whatever port you chose

Documentation and tutorials

AgentNet is using embedded documentation, so calling help(some_function_or_object) or pressing shift+tab in Ipython will yield description of what that thing is supposed to do.

A standard pipeline of AgentNet experiment can be found among examples

• Playing Atari SpaceInvaders with Convolutional NN via OpenAI Gym
  • Step-by-step explaination of what you need to do to recreate DeepMind Atari DQN
  • Written in a generic way, so that adding recurrent memory or changing learning algorithm could be done in a couple of lines
• Simple Deep Recurrent Reinforcement Learning setup
  • Most basic demo, if a bit boring. Covers the problem of learning "If X1 than Y1 Else Y2".
  • Only required if SpaceInvaders left you confused.

Demos

If you wish to get acquainted with the current library state, view some of the ./examples

• Playing Atari with Convolutional NN via OpenAI Gym
  • Can switch to any visual game thanks to their awesome interface
  • Very simplistic, non-recurrent suffering from atari flickering, etc.
• Deep Recurrent Kung-Fu training with GRUs and actor-critic
  • Uses the "Playing atari" example with minor changes
  • Trains via Advantage actor-critic (value+policy-based)
• Simple Deep Recurrent Reinforcement Learning setup
  • Trying to guess the interconnected hidden factors on a synthetic problem setup
• Stack-augmented GRU generator
  • Reproducing http://arxiv.org/abs/1503.01007 with less code
• MOAR deep recurrent value-based LR for wikipedia facts guessing
  • Trying to figure a policy on guessing musician attributes (genres, decades active, instruments, etc)
  • Using several hidden layers and 3-step Q-learning
• More to come

If you wish to join the development, we would be eager to accept your help. Current priority development anchors are maintained at the bottom of this readme.