Skip to content
Official implementation for the paper: "Shallow Updates for Deep Reinforcement Learning"
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
NeuralQLearnerShallow.lua
README.md
alg_diag.png

README.md

Shallow_Updates_for_Deep_RL

Official implementation for the paper: "Shallow Updates for Deep Reinforcement Learning"

ScreenShot

This code extends the official DQN implementation [link].

To run:

  • Follow the official isntallation isntructions.
  • Replace NeuralQLearner.lua with our code (NeuralQLearnerShallow.lua)
  • Change initenv.lua, line 14: from "require 'NeuralQLearner'" to "require 'NeuralQLearnerShallow'".

Features supported:

  • Netowrk type: DQN, and DDQN (line 155).
  • SRL method: LSTD-Q, FQI, or none (line 156).
  • Regularizer coefficient for the SRL methods (line 157).
  • SRL updates frequency (line 158).
You can’t perform that action at this time.