Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Spectral Analysis of Representational Similarity with Limited Neurons
Authors: Hyunmo Kang, Abdulkadir Canatar, SueYeon Chung
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Theoretical predictions are validated on synthetic and real datasets, offering practical strategies for interpreting neural data under finite sampling constraints. |
| Researcher Affiliation | Academia | Hyunmo Kang Department of Physics Johns Hopkins University Baltimore, MD 21218 EMAIL, Abdulkadir Canatar Center for Computational Neuroscience Flatiron Institute New York, NY 10010 EMAIL, Sue Yeon Chung Department of Physics Harvard University Cambridge, MA 02138 EMAIL |
| Pseudocode | Yes | Algorithm 1 Inferring Pop. Cross-Overlap M |
| Open Source Code | Yes | Code for all experiments is publicly available in this Github repository. |
| Open Datasets | Yes | We employ a set of publicly available neural recordings from primate visual cortex (e.g., V2) and compare these against the representations of various vision models, similarly to the methodology in [12]. |
| Dataset Splits | No | The paper discusses 'sampling a finite number of neurons' and 'neuron subsampling' from a 'full underlying population'. It also refers to 'a fixed set of stimuli of size P' and 'an artificially limited neuron count of N = 20 out of 103 neurons' for real neural recordings. However, it does not specify traditional training, validation, or test dataset splits. |
| Hardware Specification | Yes | All experiments were done using a single A100 GPU. |
| Software Dependencies | No | The paper mentions 'Mathematica' in SI.C.1 for calculations, but no version number. No other specific software with version numbers are provided for running the experiments. |
| Experiment Setup | Yes | For power-law population spectra λi = i 1 γ (used in synthetic experiments), 'We set P = 200 and N = 30' (in synthetic example), 'we will refer to SVCCA truncated to the top ten components for both X and Y as CCA' (in section 2.1), and 'artificially limited neuron count of N = 20 out of 103 neurons' (in section 5.2 for brain data). |