Sohan Seth, PhD
Research Associate
Institute for Adaptive and Neural Computation
School of Informatics
University of Edinburgh

Office: 2.29, Informatics Forum, 10 Crichton Street, Edinburgh EH8 9AB

Email: sseth(at)inf(dot)ed(dot)ac(dot)uk


I am currently a research associate at the Institute for Adaptive and Neural Computation in University of Edinburgh. I am working with Dr. Chris Williams. Before joining UE, I was a postdoctoral researcher in the statistical machine learning and bioinformatics group and Helsinki Institute for Information Technology (HIIT) in the Aalto University. I worked with Dr. Samuel Kaski and Dr. Antti Honkela. I received my PhD degree in Aug, 2011 at the University of Florida in the Electrical and Computer Engineering department, where I worked with Dr. Jose C. Principe in Computational NeuroEngineering Laboratory. 

I am from India; specifically from a place named Bally near Kolkata, West Bengal. I received my bachelor degree in 2005, in  Instrumentation and Electronics Engineering from Jadavpur University, Kolkata. After completion of my degree I worked as a Junior Research Fellow in the same institute with Dr. Rajib Bandyopadhyay and Dr. Anutosh Chatterjee for about a year before joining UF in 2006. Here I received by master degree in Electrical and Computer Engineering department in Dec. 2008. 

Currently, I am part of PROTEUS, an interdisciplinary project for developing bedside healthcare technology for critically injured patients. During my stay in Finland, I worked on information retrieval with particular focus on retrieval of experiments and retrieval of metagenomic samples. I has also been exploring probabilistic methods for matrix factorization, in particular archetypal analysis. During my PhD, I worked on nonparametric measures of independence, dependence, conditional independence, and conditional dependence, and their applications in independent component analysis, feature selection, Granger causal inference and spike train analysis. I am also interested in designing strictly positive definite kernels on spike trains for neural decoding and hypothesis testing.

You can follow me on twitter.