Skip to content

Code for Master's Project - Exploring Generalization in Text-Based Reinforcement Learning

Notifications You must be signed in to change notification settings

bryonkucharski/text-based-rl-generalization

Repository files navigation

Code for Master's Project Report: Exploring Generalization in Text-based Reinforcement Learning

Goal

Goal was to test a reinforcement learning agent's ability to generalize to unseen text-based games

Model Architecture

architecture

Example Input

Input Type Example Text
Description -= Kitchen =-If you're wondering why everything seems so normal all of a sudden, it's because you've & just shown up in the kitchen. You can see a closed fridge, which looks conventional, right there by you. You see a closed oven right there by you. Oh, great. Here's a table. Unfortunately, there isn't a thing on it. Hm. Oh well You scan the room, seeing a counter. The counter is vast. On the counter you can make out a cookbook and a knife. You make out a stove. Looks like someone's already been here and taken everything off it, though. Sometimes, just sometimes, TextWorld can just be the worst.
Inventory You are carrying: a banana
Recipe Recipe #1 - Gather all following ingredients and follow the directions to prepare this tasty meal. Ingredients: banana Directions: chop the banana fry the banana prepare meal
Previous Action take banana from counter
Feedback You take the banana from the counter. Your score has just gone up by one point

Results

Plots across four Levels of difficulties (left to right)

Normalized score is y axis

Training One Game at a time

single_game

Evaluation Results across 20 evaluation games

Evaluation

About

Code for Master's Project - Exploring Generalization in Text-Based Reinforcement Learning

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages