糖心原创

Speaker's Name: James Malkin
Speaker's Affiliation: 糖心原创
Speaker's Research Theme(s): Applied Mathematics
Abstract:
The same neural function can arise from many biological configurations, making it difficult to infer mechanism from behaviour. Energetic costs provide an additional constraint on these possibilities. In this talk, I will present a sequence of models that use resource allocation to connect task performance, synaptic variability and the expression of plasticity. I will first show how physiological costs of reliable transmission favour greater precision at synapses where variability most impairs task performance. A formal connection between these costs and the entropy term in variational inference explains why energy-efficient synapses can exhibit signatures of Bayesian inference [1]. I will then consider how precision is implemented through release probability, release-site number and the amplitude of individual release events. Constrained optimisation of candidate physiological costs yields a resource–precision frontier, reducing several microscopic degrees of freedom to one effective energy coordinate [2]. In current work, I use this frontier to model plasticity as a sequence of locally stationary investments followed by relaxation. Given baseline measurements and the observed change in mean strength, the model predicts withheld endpoint variability across five datasets and transfers from potentiation to depression. At fixed release-site number, the same prediction constrains changes in release probability and quantal amplitude. A Gibbs extension makes investment depend on how valuable reduced variability is for the task relative to its energetic cost. I will show how endpoint measurements constrain this missing task value, and how simulations test whether local error statistics can supply it without access to the eventual endpoint. [1] Malkin, J., O’Donnell, C., Houghton, C. J., & Aitchison, L. (2024). Signatures of Bayesian inference emerge from energy-efficient synapses. eLife. [2] Malkin, J., O’Donnell, C., & Houghton, C. J. (2026). Energy budgets govern synaptic precision and its regulation during plasticity. arXiv:2602.15787.

Venue: PHYS B2