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..
Inference with correlated priors using sisters cells
Authors: Sina Tootoonian, Andreas Schaefer
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Using simulations, we demonstrate the efficacy of such priors for inference in noisy environments and compare the inference dynamics to those experimentally observed. |
| Researcher Affiliation | Academia | Sina Tootoonian and Andreas T. Schaefer Sensory Circuits and Neurotechnology Laboratory The Francis Crick Institute London, UK [sina.tootoonian|andreas.schaefer]@crick.ac.uk |
| Pseudocode | No | The paper describes dynamic equations (Eqn. 2, 3, 4, 5) but does not present a distinct pseudocode block or algorithm box. |
| Open Source Code | Yes | All code and data required to reproduce the results in this work are provided in the supplementary material. Linking to the github repository here would break anonymity, but a link will be provided should the paper be accepted. |
| Open Datasets | Yes | To quantitatively evaluate the quality of these estimates, we consulted a publicly available database of 214 essential oils and their monomolecular components [13]. |
| Dataset Splits | No | The paper describes generating random noisy corruptions across 5 trials for evaluation and defines parameters for simulated data, but it does not specify train/test/validation splits for a pre-existing dataset. |
| Hardware Specification | Yes | All simulations were carried out in python version 3.9 running on a mid-2015 2.8 GHz Intel Core i7 Mac Book Pro, and all individual simulations ran in about one minute or less. |
| Software Dependencies | Yes | All simulations were carried out in python version 3.9... Optimizations were performed using the pymanopt package using its Steepest Descent optimizer. ...by solving the convex optimization directly, using the cvxpy python package with the SCS solver. |
| Experiment Setup | Yes | The standard deviation of the receptor noise and that of inference were σn = 0.5 and σinf = 20, respectively, unless otherwise noted. In all simulations the individual prior on the latents was the elastic net ϕi(xi) = βxi + γi where the ℓ1 parameter β = 0.1, unless otherwise noted, and the ℓ2 parameter γi was set per unit so that its sum with the corresponding diagonal term coming from the correlated prior was 0.1. ... The integration time constants of the mitral cells and latent feature units were τmc = 50 msec. and τgc = 100 msec., respectively. These were selected because they were biologically realistic and gave smooth dynamics that converged within respiration time. We used first-order Euler integration with a step size of 200µsec to integrate the dynamics. |