Layered State Discovery for Incremental Autonomous Exploration

Authors: Liyu Chen, Andrea Tirinzoni, Alessandro Lazaric, Matteo Pirotta

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical Evaluations We implemented our LASD algorithm and evaluated it empirically. Implementations can be found in https://github.com/lchenat/AX_exp. We manually tune the values of some parameters such as nmin and λ to boost the empirical performance, and then conducted experiments on a 4x4 Grid World environment.
Researcher Affiliation Collaboration 1University of Southern California 2Meta. Correspondence to: Liyu Chen <liyuc@usc.edu>.
Pseudocode Yes Algorithm 1: Layer-Aware State Discovery (LASD)
Open Source Code Yes Implementations can be found in https://github.com/lchenat/AX_exp.
Open Datasets No We manually tune the values of some parameters such as nmin and λ to boost the empirical performance, and then conducted experiments on a 4x4 Grid World environment. The learner has 5 actions in this environment: moving towards one of the four directions by a grid or reset to s0 (the upper left corner). When the learner takes a directional action, it has probability 0.9 of moving towards the corresponding direction, and 0.1 probability of randomly moving towards one of the four directions.
Dataset Splits No The paper does not specify any training/validation/test dataset splits, as it conducts experiments in a simulated 4x4 Grid World environment, which is not typically split like a dataset.
Hardware Specification No The paper mentions running empirical evaluations but does not provide specific details about the hardware used (e.g., CPU/GPU models, memory, or cloud instance types).
Software Dependencies No The paper provides a link to its implementation but does not specify any software dependencies with version numbers (e.g., programming language versions, library versions).
Experiment Setup Yes We manually tune the values of some parameters such as nmin and λ to boost the empirical performance, and then conducted experiments on a 4x4 Grid World environment. ... We run LASD on Grid World with L = 4, ϵ = 0.01, and δ = 0.001.