Learning What We Know and Knowing What We Learn - Gaussian Process Priors for Neural Data Analysis

Explore the intricacies of Gaussian process priors for neural data analysis in this comprehensive lecture. Delve into the importance of latent variable models, Bayesian inference, and covariance kernels in neural data analysis. Learn about factor analysis, spectral mixture kernels, and margin likelihood through practical examples and Colab notebooks. Discover the challenges and limitations of data-driven approaches, and examine the results of Bayesian GPFA. Gain insights from additional resources, including papers on Gaussian process factor analysis with dynamical structure and extensions to non-Euclidean manifolds. Understand how these techniques can be applied to real-world scenarios, such as analyzing hippocampal encoding in evidence accumulation tasks.

Introduction
Why should we use latent variable models
Fitting variable models
Data hungry
Simple regression
Bayesian inference
Covariance
Covariance kernels
Spectral mixture kernels
Margin likelihood
Correlation kernel
Factor analysis
Collab notebook
Challenges
Bayesian GPFA
Data limitations
Results