Origin of the spike-timing–dependent plasticity rule
1 Department of Global Medical Science, Sungshin Women's University - Seoul 01133, Korea
2 Department of Physics and Astronomy and Center for Theoretical Physics, Seoul National University Seoul 08826, Korea
Received: 12 February 2016
Accepted: 20 August 2016
A biological synapse changes its efficacy depending on the difference between pre- and post-synaptic spike timings. Formulating spike-timing–dependent interactions in terms of the path integral, we establish a neural-network model, which makes it possible to predict relevant quantities rigorously by means of standard methods in statistical mechanics and field theory. In particular, the biological synaptic plasticity rule is shown to emerge as the optimal form for minimizing the free energy. It is further revealed that maximization of the entropy of neural activities gives rise to the competitive behavior of biological learning. This demonstrates that statistical mechanics helps to understand rigorously key characteristic behaviors of a neural network, thus providing the possibility of physics serving as a useful and relevant framework for probing life.
PACS: 87.85.dm – Physical models of neurophysiological processes / 87.18.Sn – Neural networks and synaptic communication / 87.19.lw – Plasticity
© EPLA, 2016