Valerio Mante

Wednesday 30th May 2018

Time:4.00pm

Ground Floor Seminar Room

25 Howland Street, London, W1T 4JG

Nearest-neighbor graphs reveal multiple time-scales of learning in Zebra finches

Thanks to advances in experimental techniques, behavioral and neural data sets in neuroscience are becoming ever larger. Inferring structure in such large data sets is challenging, but concurrent increases in the speed of computers have opened the way to novel, computationally intensive analysis approaches. I will present such an approach that is based on the characterization of “nearest-neighbor relations" in the data. Our method is non-parametric and makes minimal assumptions about the nature of the data, and is thus applicable to any large, dense data set. I will demonstrate the power of this method on large-scale recordings of vocalization in songbirds, where it provides a unified description of the dynamics of song learning over many time-scales, from weeks to days, hours, and down to individual syllable renditions.