Hierarchies of Reward Machines

Authors: Daniel Furelos-Blanco, Mark Law, Anders Jonsson, Krysia Broda, Alessandra Russo

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

Reproducibility Variable Result LLM Response
Research Type Experimental 6. Experimental Results
Researcher Affiliation Collaboration 1Imperial College London, UK 2ILASP Limited, UK 3Universitat Pompeu Fabra, Spain.
Pseudocode Yes We refer the reader to Appendix B.2 for the pseudo-code and step-by-step examples.
Open Source Code Yes The code is available at https://github. com/ertsiger/hrm-learning.
Open Datasets Yes The CRAFTWORLD domain (cf. Figure 1a) is used as a running example. In this domain, the agent ( ) can move forward or rotate 90 , staying put if it moves towards a wall. Locations are labeled with propositions from P = { , , , , , , , , , }. ... WATERWORLD (Karpathy, 2015; Sidor, 2016; Toro Icarte et al., 2018) consists of a 2D box containing 12 balls of 6 different colors (2 per color) each moving at a constant speed in a fixed direction.
Dataset Splits No The paper describes using multiple runs and instances for evaluation and a curriculum learning approach for tasks, but it does not specify explicit train/validation/test dataset splits with percentages or counts for reproduction in a traditional supervised learning sense. It focuses on task-instance pairs and average returns across episodes and runs.
Hardware Specification Yes All timed experiments ran on 3.40GHz Intel Core i7-6700 processors, while non-timed experiments have also run on 2.90GHz Intel Core i7-10700, 4.20GHz Intel Core i7-7700K, and 3.20GHz Intel Core i7-8700 processors.
Software Dependencies No The paper mentions software components and algorithms like 'Deep Q-networks', 'Double DQNs', and 'RMSprop', but it does not provide specific version numbers for software libraries, frameworks, or programming languages (e.g., PyTorch 1.x, Python 3.x).
Experiment Setup Yes Table 2: List of hyperparameters and their values.