Synaptic plasticity as Bayesian inference

TL;DR

This paper proposes that synapses perform Bayesian inference by representing uncertainty in weights, adjusting learning rates accordingly, and linking uncertainty to variability in PSP size, providing a normative framework for understanding synaptic plasticity.

Contribution

It introduces a novel Bayesian inference model for synaptic plasticity, explaining variability and adaptive learning in neural synapses.

Findings

Synapses encode uncertainty in weights as error bars.

Uncertainty influences synaptic learning rates.

Variability in PSP size correlates with synaptic uncertainty.

Abstract

Learning, especially rapid learning, is critical for survival. However, learning is hard: a large number of synaptic weights must be set based on noisy, often ambiguous, sensory information. In such a high-noise regime, keeping track of probability distributions over weights is the optimal strategy. Here we hypothesize that synapses take that strategy; in essence, when they estimate weights, they include error bars. They then use that uncertainty to adjust their learning rates, with more uncertain weights having higher learning rates. We also make a second, independent, hypothesis: synapses communicate their uncertainty by linking it to variability in PSP size, with more uncertainty leading to more variability. These two hypotheses cast synaptic plasticity as a problem of Bayesian inference, and thus provide a normative view of learning. They generalize known learning rules, offer an explanation for the large variability in the size of post-synaptic potentials, and make falsifiable experimental predictions.