Deep probabilistic model synthesis enables unified modeling of whole-brain neural activity across individual subjects

TL;DR

This paper introduces Deep Probabilistic Model Synthesis (DPMS), a machine learning framework that combines data across multiple system instances, such as individual brains, to improve modeling of neural activity.

Contribution

DPMS leverages auxiliary system properties and variational inference to unify data across instances, enabling synthesis of diverse model types and improved performance over single-instance models.

Findings

DPMS improves modeling accuracy on synthetic data.

DPMS enhances neural activity modeling across zebrafish brains.

The framework is applicable to various model classes.

Abstract

Many disciplines need quantitative models that synthesize experimental data across multiple instances of the same general system. For example, neuroscientists must combine data from the brains of many individual animals to understand the species' brain in general. However, typical machine learning models treat one system instance at a time. Here we introduce a machine learning framework, deep probabilistic model synthesis (DPMS), that leverages system properties auxiliary to the model to combine data across system instances. DPMS specifically uses variational inference to learn a conditional prior distribution and instance-specific posterior distributions over model parameters that respectively tie together the system instances and capture their unique structure. DPMS can synthesize a wide variety of model classes, such as those for regression, classification, and dimensionality reduction, and we demonstrate its ability to improve upon single-instance models on synthetic data and whole-brain neural activity data from larval zebrafish.