Auxiliary Counting Variables for Optimal Decoding...

Περίληψη:
Hidden Markov Models (HMMs) are one of the most popular methods for
sequence analysis in statistics and signal processing including genetics
and genomics where they have found widespread use. However, posterior
summaries of dependencies found from modelling are poorly developed.
Motivated by applications in cancer genomics we present new
computational algorithms that can reveal interesting structure in the
data. In particular we present a linear time method for deriving the
most probable hidden state sequence conditional on a user defined number
of state-transitions, possibly of a particular type. This provides
optimal K-segment partitioning of the data. The methods, while developed
for applications in cancer genomics, are generic and importantly can be
applied retrospectively to HMMs already fitted to data.

Λίγα λόγια για τον ομιλητή:
Michalis Titsias received a Diploma in Informatics from the University
of Ioannina, Greece, in 1999, an MSc degree also from the University of
Ioannina, in 2001, and a PhD degree from the School of Informatics,
University of Edinburgh, in 2005. From October 2007 to July 2011, he
worked as a research associate in the machine learning and optimization
research group at the School of Computer Science of the University of
Manchester, while from November 2011 to September 2012 he worked as a
postdoctoral research scientist in statistical cancer genomics at the
Wellcome Trust Centre for Human Genetics and the Department of
Statistics at the University of Oxford. Starting form 2012, he is a
Lecturer in the Department of Informatics of the Athens University of
Economics and Business. His research interests include applied Bayesian
statistics, machine learning, bioinformatics and statistical cancer
genomics.