Gain-Adaptive Recurrent Networks Explain Fast Sensory Adaptation and Efficient Coding

A new study published in Nature Communications proposes a unified theoretical model explaining how sensory systems rapidly adapt to changing environmental statistics. Researchers from Harvard University, the University of Zurich, and the Technical University of Munich developed a gain-adaptive recurrent sensory network model. This model demonstrates that gain modulation optimizes an efficient-coding objective by balancing accuracy and spiking costs. The framework successfully reconciles two seemingly contradictory phenomena observed in neural tuning curves: 'adapter repulsion,' seen after repeated stimulus presentations, and 'prior attraction,' expected under broad distributions. By simulating the propagation of modulated gains throughout the network, the model accounts for subtle adapter-repulsion effects under peaked priors while predicting fast prior attraction for broader distributions. The authors provide supporting behavioral evidence for these predictions. This research offers a mechanistic understanding of how neural circuits achieve fast, efficient coding, thereby supporting adaptive behavior in dynamic sensory environments. The findings bridge gaps in previous efficient-coding models and clarify the underlying mechanisms of sensory processing and neural encoding.