A Classification of $G$-invariant Shallow Neural Networks

Authors: Devanshu Agrawal, James Ostrowski

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Using a code implementation, we enumerate the G-SNN architectures for some example groups G and visualize their structure. Using our code implementation, we enumerated all irreducible G-SNN architectures for one permutation representation of every group G, |G| 8, up to isomorphism. We report the number of architectures, broken down by type, for each group in Table 1 (see Supp. C.1; a discussion is included there as well). Script execution time for each group was under 2 seconds.
Researcher Affiliation Academia Devanshu Agrawal & James Ostrowski Department of Industrial and Systems Engineering University of Tennessee Knoxville, TN 37996 dagrawa2@vols.utk.edu, jostrows@utk.edu
Pseudocode No The paper describes theoretical concepts, proofs, and a classification algorithm but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks.
Open Source Code Yes Code for our implementation and for reproducing all results in this paper is available at: https://github.com/dagrawa2/gsnn_classification_code.
Open Datasets No The paper focuses on theoretical classification and enumeration of neural network architectures, not on training models with specific datasets. Therefore, it does not mention the use of any publicly available or open datasets for training or evaluation.
Dataset Splits No The paper is primarily theoretical, focusing on the classification and enumeration of G-SNN architectures. It does not involve training machine learning models on datasets that would require explicit training, validation, or test splits.
Hardware Specification No The paper does not provide any specific details regarding the hardware used for running the code implementation, such as GPU models, CPU types, or cloud computing resources.
Software Dependencies Yes We implemented the enumeration algorithm using a combination of GAP and Python; our implementation currently supports, in principle, all finite permutation groups G < P(m). and GAP. GAP Groups, Algorithms, and Programming, Version 4.11.1. https://www.gap-system.org, 2021.
Experiment Setup Yes To visualize these architectures, we set w = [1, 0] , a = 1, b = 0.5, c = 0, and d = 0 in Thm. 4 (b).