Reshaping Neural Representation via Associative, Presynaptic Short-Term Plasticity
Short-term synaptic plasticity (STP) is often regarded as a presynaptic filter of spikes, independent of postsynaptic activity. Recent experiments, however, indicate an associative STP that depends on pre- and postsynaptic coactivation. We develop a normative, information-theoretic theory of associative STP. Extending Fisher-information-based learning to Tsodyks-Markram synapses, we derive learning rules for baseline weight and release probability that maximize stimulus information under resource constraints. The rules split into a postsynaptic term tracking local firing and a presynaptic, phase-advanced term that selectively detects stimulus onset. For slowly varying inputs, this onset sensitivity favors anti-causal connectivity and enhances response offset during drive and reverse replay after drive removal in recurrent circuits. Linear-response analysis shows that STP yields frequency-dependent phase selectivity and that release-probability constraints tune temporal asymmetry. These results identify release-probability plasticity as a principled substrate for rapidly reconfigurable temporal coding.