Learning identifiable and interpretable latent models of high-dimensional neural activity using pi-VAE

Authors: Ding Zhou, Xue-Xin Wei

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

Reproducibility Variable Result LLM Response
Research Type Experimental We validate pi-VAE using synthetic data, and apply it to analyze neurophysiological datasets from rat hippocampus and macaque motor cortex. We demonstrate that pi-VAE not only fits the data better, but also provides unexpected novel insights into the structure of the neural codes.
Researcher Affiliation Academia Ding Zhou Department of Statistics Columbia University dz2336@columbia.edu Xue-Xin Wei Department of Neuroscience UT Austin weixx@utexas.edu
Pseudocode No The paper describes the inference procedure in text but does not provide pseudocode or a clearly labeled algorithm block.
Open Source Code No The paper does not provide any explicit statement about releasing source code or a link to a code repository for their methodology.
Open Datasets Yes We have applied pi-VAE to analyze two electrophysiology datasets, each has more than 100 simultaneously recorded neurons when the animals were performing behavioral tasks. [...] a public rat s hippocampus dataset (27; 26)1. http://crcns.org/data-sets/hc/hc-11
Dataset Splits Yes We randomly split the dataset into 24 batches, where each batch contains at least one trial for each direction. We randomly split them into training, validation and test data (20, 2, 2 batches). [...] We randomly split them into training, validation and test data (68, 8, 8 laps).
Hardware Specification No The paper does not explicitly describe the hardware used for running its experiments, such as specific GPU/CPU models or memory details.
Software Dependencies No The paper mentions using the Adam optimizer and setting values to recommendations, but does not provide specific version numbers for any software dependencies or libraries (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup Yes We binned the ensemble spike activities into 50ms bins. [...] We used the spike activities as observation x, and the reaching direction as the discrete labels u. [...] We fit 4-dimensional latent models to the data based on pi-VAE and VAE. [...] We binned the ensemble spike activities into 25ms bins. [...] We fit 2-dimensional latent models to the data for both pi-VAE and VAE. [...] Adam optimizer (36) with learning rate equal to 5 10 4, and other values were set to the recommendation values for all the experiments in this paper.