Synchrony-Based State Representation for Classification by Liquid State Machines

The Liquid State Machine (LSM) models usually ignore the influence of the liquid state representation on performance, with the assumption that the latter depends only on the readout circuit. The typical decoding of the liquid’s spike trains is achieved with spike rate-based vectors that are input into the readout circuit. This occults the spike timing, a central aspect of biological neural coding, with potentially detrimental consequences on performance. We propose a model of liquid state representation that builds the feature vectors from the temporal information about the spike trains, hence using spike synchrony instead of rate. Using Poisson-distributed spike trains in noisy conditions, we show that such model outperforms a rate-only model in distinguishing spike train pairs, regardless of the frequency chosen to sample the liquid state or the noise level. In the same vein, we suggest a synchrony-based measure of the Separation Property (SP), a core feature of LSMs regarding classification performance, for a more robust and biologically plausible interpretation.