Abstract:Modern recording techniques enable neuroscientists to simultaneously study neural activity across large populations of neurons, with capturing predictor-dependent correlations being a fundamental challenge in neuroscience. Moreover, the fact that input covariates often lie in restricted subdomains, according to experimental settings, makes inference even more challenging. To address these challenges, we propose a set of nonparametric mean-covariance regression models for high-dimensional neural activity with restricted inputs. These models reduce the dimensionality of neural responses by employing a lower-dimensional latent factor model, where both factor loadings and latent factors are predictor-dependent, to jointly model mean and covariance across covariates. The smoothness of neural activity across experimental conditions is modeled nonparametrically using two Gaussian processes (GPs), applied to both loading basis and latent factors. Additionally, to account for the covariates lying in restricted subspace, we incorporate graph information into the covariance structure. To flexibly infer the model, we use an MCMC algorithm to sample from posterior distributions. After validating and studying the properties of proposed methods by simulations, we apply them to two neural datasets (local field potential and neural spiking data) to demonstrate the usage of models for continuous and counting observations. Overall, the proposed methods provide a framework to jointly model covariate-dependent mean and covariance in high dimensional neural data, especially when the covariates lie in restricted domains. The framework is general and can be easily adapted to various applications beyond neuroscience.

Large-scale Collaborative Prediction Using a Nonparametric Random Effects Model.

Structured Sparse Linear Model For Social Trust Prediction

Distributed Learning of Predictive Structures from Multiple Tasks over Networks

Parallel integrative learning for large-scale multi-response regression with incomplete outcomes

Collaborative Performance Prediction for Large Language Models

Large-Scale Dynamic Predictive Regressions

Multivariate Spatial-temporal Prediction on Latent Low-dimensional Functional Structure with Non-stationarity

Predicting Cross-Domain Collaboration Using Multi-Task Learning

Prediction under Latent Subgroup Shifts with High-Dimensional Observations

Nonparametric Covariance Regression for Massive Neural Data on Restricted Covariates via Graph

Scalable Multilabel Prediction via Randomized Methods

Model Free Prediction with Uncertainty Assessment

Prediction-driven collaborative emergency medical resource allocation with deep learning and optimization

Adaptive Ensemble Learning of Spatiotemporal Processes with Calibrated Predictive Uncertainty: A Bayesian Nonparametric Approach

Fast Nonparametric Matrix Factorization For Large-Scale Collaborative Filtering

GPU Accelerated Estimation of a Shared Random Effect Joint Model for Dynamic Prediction

A Self-Organizing Method for Predictive Modeling with Highly-Redundant Variables

A Random-effects Approach to Regression Involving Many Categorical Predictors and Their Interactions

Learning to Predict Parameter for Unseen Data

Collaboration Prediction in Heterogeneous Academic Network with Dynamic Structure and Topic.

Collaborative Online RUL Prediction of Multiple Assets With Analytically Recursive Bayesian Inference