Neuroscience Group Meeting

Topic: Clustering time series in the brain: connections to information theory, canonical correlations, and cepstra

Sensory information reaches the brain as a stream with non-trivial correlations across time. In a generative model, these correlations can be seen as the result of a dynamical system acting on a white noise source signal. Learning the parameters describing this system enables a variety of applications, from detecting changes in the input dynamics to inferring dynamical rules in the environment. I’ll go over topics from information theory, canonical correlation analysis, and cepstral analysis that can be used to characterize and cluster signals based on the dynamical systems that generated them. I’ll then present several online algorithms that can be used to perform clustering and system identification in a streaming setting, without needing access to the whole signal at once. Finally, I’ll talk about prospects for implementing these or related algorithms in biologically plausible neural networks.