AUTHOR=Schönfeld Fabian , Wiskott Laurenz 

TITLE=Modeling place field activity with hierarchical slow feature analysis

JOURNAL=Frontiers in Computational Neuroscience

VOLUME=Volume 9 - 2015

YEAR=2015

URL=https://www.frontiersin.org/journals/computational-neuroscience/articles/10.3389/fncom.2015.00051

DOI=10.3389/fncom.2015.00051

ISSN=1662-5188

ABSTRACT=In this paper we present six experimental studies from the literature on hippocampal place cells and replicate their main results in a computational framework based on the principle of slowness. Each of the chosen studies first allows rodents to develop stable place field activity and then examines a distinct property of the established spatial encoding, namely adaptation to cue relocation and removal; directional firing activity in the linear track and open field; and results of morphing and stretching the overall environment. To replicate these studies we employ a hierarchical Slow Feature Analysis (SFA) network. SFA is an unsupervised learning algorithm extracting slowly varying information from a given stream of data, and hierarchical application of SFA allows for high dimensional input such as visual images to be processed efficiently and in a biologically plausible fashion. Training data for the network is produced in ratlab, a free basic graphics engine designed to quickly set up a wide range of 3D environments mimicking real life experimental studies, simulate a foraging rodent while recording its visual input, and training & sampling a hierarchical SFA network.