A typical machine learning model looks at one object at a time: one brain, one patient, one engine. But biologists need the big picture: how the brain of a species works as a whole, not just that of a particular fish. The new method — deep probabilistic model synthesis (DPMS)—learns from many instances at once, like a chef who tastes dozens of apples to understand the flavor of the variety, not just one fruit.

The approach applies to various model types, from regression to dimensionality reduction. On synthetic data and brain activity recordings from zebrafish larvae, DPMS outperformed models trained on a single instance. This is a step toward understanding general principles of complex systems from many of their manifestations.