Efficient estimation of neural tuning during naturalistic behavior

Authors: Edoardo Balzani, Kaushik Lakshminarasimhan, Dora Angelaki, Cristina Savin

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental When applied to neural recordings from monkeys performing a virtual reality spatial navigation task, P-GAM reveals mixed selectivity and preferential coupling between neurons with similar tuning. We demonstrate the efficiency of this procedure using artificial data, showing that P-GAM outperforms standard GLMs.
Researcher Affiliation Academia Edoardo Balzani Center for Neural Science New York University New York, NY, 10003 eb162@nyu.edu Kaushik Lakshminarasimhan Center for Theoretical Neuroscience Columbia University New York, NY, 10027 jl5649@columbia.edu Dora E. Angelaki Center for Neural Science New York University New York, NY, 10003 da93@nyu.edu Cristina Savin Center for Neural Science Center for Data Science New York University New York, NY, 10003 cs5360@nyu.edu
Pseudocode No The paper describes iterative procedures in numbered list format within the text, but does not include explicitly labeled "Pseudocode" or "Algorithm" blocks/figures.
Open Source Code Yes 1Code available at: https:/github.com/Balzani Edoardo/PGAM.
Open Datasets Yes When applied to neural recordings from monkeys performing a virtual reality spatial navigation task, P-GAM recovers known features of the neural code, in particular mixed selectivity[15] and structured noise correlations [16, 17, 18]. [3] Kaushik J. Lakshminarasimhan, Eric Avila, Erin Neyhart, Gregory C. De Angelis, Xaq Pitkow, and Dora E. Angelaki. Tracking the Mind s Eye: Primate Gaze Behavior during Virtual Visuomotor Navigation Reflects Belief Dynamics. Neuron, 106(4):662 674.e5, May 2020.
Dataset Splits No The paper mentions using cross-validation for hyperparameter optimization and model comparison ("optimized using a grid search, with the cross-validated pseudo-r2 as optimization objective"), but does not provide specific training/validation/test dataset split percentages or sample counts.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions using the "statsmodels python library [29]" but does not provide a specific version number for it or any other software dependencies.
Experiment Setup Yes The order of the splines and the knots locations are model hyperparameters that could be optimized by cross-validation; in practice, we use cubic splines (m = 4) and manually choose knots that reasonably cover the input range. We calibrated the parameters of the ground truth model to a relatively realistic setting (30min of data, sampled in 6ms bins; 5Hz mean firing rates, see source code for details). A typical session lasts about 90min, with spike counts measured in 6ms time bins; the analysis presented here includes 30 sessions from one animal.