bgLeon/SysGen-DRL-TL

Code from the ALOE paper titled Agent, do you see it now? systematic generalisation in deep reinforcement learning

Systematic Generalisation of abstract operators

Source code of Agent, do you see it now? systematic generalisation in deep reinforcement learning @ ALOE ICLR 2022 and 'Systematic Generalisation of Temporal Tasks through Deep Reinforcement Learning' @ 'ALA AAMAS 2021'.

Installation instructions

This repository requires Python3.6 with three libraries: numpy, tensorflow (V2.5), and gym.

Running example

python run_experiment.py --visual --mapType=minigrid --syntax=TTL1 --network=PrediNet --mem --seedB=0 --batch_size=512 --cnn_type=cnn

This runs the SA from the main document in the minigrid benchmark.

run_experiment.py contains instructions for each command