Point process models for sequence detection in high-dimensional neural spike trains

Authors: Alex Williams, Anthony Degleris, Yixin Wang, Scott Linderman

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

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate these advantages on experimental recordings from songbird higher vocal center and rodent hippocampus. We evaluate model performance by computing the log-likelihood assigned to held-out data. Partitioning the data into training and testing sets must be done somewhat carefully.
Researcher Affiliation Academia Alex H. Williams Department of Statistics Stanford University Stanford, CA 94305 ahwillia@stanford.edu Anthony Degleris Department of Electrical Engineering Stanford University Stanford, CA 94305 degleris@stanford.edu Yixin Wang Department of Statistics Columbia University New York NY 10027 yixin.wang@columbia.edu Scott W. Linderman Department of Statistics Stanford University Stanford, CA 94305 scott.linderman@stanford.edu
Pseudocode No The paper describes the algorithm steps in paragraph form, particularly in Section 3 'Collapsed Gibbs Sampling for Neyman-Scott Processes', but does not provide a formally structured pseudocode or algorithm block.
Open Source Code Yes Our open-source implementation is available at: https://github.com/lindermanlab/PPSeq.jl.
Open Datasets Yes These data are available at http://github.com/Fee Lab/seq NMF; originally published in [4]. These data are available at http://crcns.org/data-sets/hc/hc-11; originally published in [56].
Dataset Splits Yes Partitioning the data into training and testing sets must be done somewhat carefully we cannot withhold time intervals completely (as in fig. 2A, top) or else the model will not accurately predict latent sequences occurring in these intervals; likewise, we cannot withhold individual neurons completely (as in fig. 2A, middle) or else the model will not accurately predict the response parameters of those held out cells. Thus, we adopt a speckled holdout strategy [54] as diagrammed at the bottom of fig. 2A. We treat held-out spikes as missing data and sample them as part of the MCMC algorithm. We performed a large cross-validation sweep over 2,000 random hyperparameter settings for this dataset (see Supplement F.2).
Hardware Specification Yes Figure 2G shows that our Julia [55] implementation can fit a recording of 120 hippocamapal neurons with hundreds of thousands of spikes in a matter of minutes, on a 2017 Mac Book Pro (3.1 GHz Intel Core i7, 4 cores, 16 GB RAM).
Software Dependencies No The paper mentions 'Julia [55] implementation' but does not provide specific version numbers for Julia or any other software libraries or dependencies used in the experiments.
Experiment Setup Yes We performed a large cross-validation sweep over 2,000 random hyperparameter settings for this dataset (see Supplement F.2). This confirmed that models with R = 2 sequence performed well in terms of heldout performance (fig. 5D). Interestingly, despite variability in running speeds, this same analysis did not show a consistent benefit to including larger time warping factors into the model (fig. 5E). Higher performing models were characterized by larger sequence amplitudes, i.e. larger values of E[Ak] = α/β, and smaller background rates, i.e. smaller values of λ . Other parameters had less pronounced effects on performance.