Human-Readable Neuro-Fuzzy Networks from Frequent Yet Discernible Patterns in Reward-Based Environments

Authors: John Wesley Hostetter, Adittya Soukarjya Saha, Md Mirajul Islam, Tiffany Barnes, Min Chi

IJCAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluated on Cart Pole [Barto et al., 1983], Mountain Car [Moore, 1990], and Two-Link Arm [Sutton, 1995].
Researcher Affiliation Academia John Wesley Hostetter , Adittya Soukarjya Saha , Md Mirajul Islam , Tiffany Barnes and Min Chi North Carolina State University EMAIL
Pseudocode No The paper describes the methods in detailed prose and mathematical formulations but does not include any explicitly labeled pseudocode or algorithm blocks.
Open Source Code Yes Sample code demonstrating FYD is public (MIT license) [Hostetter, 2025].1 1https://github.com/john Hostetter/IJCAI-2025-FYD
Open Datasets No Training data was collected by a DNN using DQL and experience replay while solving the given environment online.
Dataset Splits No During each run, the amount of data available for offline training gradually increased to show how conditions behave as more data is provided.
Hardware Specification No The paper does not explicitly mention any specific hardware details such as GPU/CPU models, memory, or cloud computing resources used for running the experiments.
Software Dependencies No All methods were optimized by Adam [Kingma and Ba, 2014].
Experiment Setup Yes Shared parameters were identical: α = 1.0, γ = 0.99, learning rate η = 3 10 4, and the batch size was 32. For FYD or CEW, CLIP used κ = 0.2, ϵ = 0.6, and ECM s distance threshold, Dthr, was 0.1. Training data was collected by a DNN using DQL and experience replay while solving the given environment online.