Navigation

Researchers at UCD

researcher

Michael Salter-Townshend

Postdoc Research Fellow Lvl Ii

School Of Computer Science & Informatics
Belfield Office Park - 8
Belfield
Dublin 4

Tel: +353 1 716
Email: michael.salter-townshend@ucd.ie

Biography

I am a research statistician working in the Clique group in UCD on the analysis of network data. This includes, but is not limited to, social networks. I use approximate Bayesian inference methods such as Variational Bayes. I'm interested in applying rich statistical models to a wide variety of network datasets and in developing efficient, scalable algorithms for this purpose.

My Ph.D. thesis concerned the modelling of multivariate, zero-inflated, spatial counts and proportions data as part of a Palaeoclimate reconstruction project. I used traditional Bayesian inference methods such as MCMC and new Gaussian approximation techniques such as the Integrated Nested Laplace Approximation. This is an example of a large scale inverse problem. 

Professional

         

Conference Contributions

Thomas Brendan Murphy and Michael Salter-Townshend; (2009) Variational Bayesian Inference for the Latent Position Cluster Model. [Chaired Session], Statistical Methods for the Analysis of Network Data in Practice, UCD , 15-JUN-09 - 17-JUN-09.
Michael Salter-Townshend; (2010) Variational Bayesian Inference for the Latent Position Cluster Model. [Chaired Session], Workshop on Modeling and Computation for Social Networks, Whistler, BC, Canada , 26-JUN-10 - 27-JUN-10.
Michael Salter-Townshend; (2007) Gaussian Approximation Techniques. [Poster Presentation], Conference of Applied Statistics Ireland, Dundrum, Tipperary , 15-JUN-07 - 16-JUN-07.
Michael Salter-Townshend; (2006) Fast Approximations to High Dimensional Response Surfaces. [Oral Presentation], High Performance Computing in Statistical Inference, Trinity College Dublin , 17-AUG-06 - 17-AUG-06.
Michael Salter-Townshend; (2006) Zero-Inflation of Compositional Data. [Oral Presentation], 21 st International Workshop on Statistical Modelling, NUI Galway , 19-JUL-06 - 19-JUL-06.
Michael Salter-Townshend; (2006) Additive Modelling of Response Surfaces. [Poster Presentation], Conference of Applied Statistics in Ireland, Tralee , 17-JUN-06 - 19-JUN-06.
Michael Salter-Townshend; (2006) Zero-Inflation of Compositional Data. [Oral Presentation], International Conference of the Royal Statistical Society, Queen's University Belfast , 30-SEP-06 - 30-SEP-06.
Michael Salter-Townshend; (2006) Additive Modelling of Response Surfaces. [Oral Presentation], Young Statisticians Meeting, Edinburgh , 12-FEB-06 - 12-FEB-06.
Michael Salter-Townshend; (2005) Temporal Smoothing. [Poster Presentation], Conference of Applied Statistics in Ireland, Enniskillen , 18-JUN-05 - 20-JUN-05.
   

Education

Year 2004 Institution: Trinity College Dublin
Qualification: MSc Subject: High Performance Computing
Year 2003 Institution: Trinity College Dublin
Qualification: BA (Mod) Subject: Computational Physics
Year 2008 Institution: Trinity College Dublin
Qualification: PhD Subject: Statistics

Languages

Irish:
French:
       

Publications

 

Book Chapters

Salter-Townshend, M. and Murphy, T.B. (2013) 'Sentiment Analysis of Online Media' In: Lausen, B., van del Poel, D. and Ultsch, A (eds). Algorithms from and for Nature and Life. Heidelberg: Springer. Link to full text [Details]
 

Peer Reviewed Journals

Salter-Townshend, M; Haslett, J (2012) 'Fast Inversion of a Flexible Regression Model for Multivariate, Zero-Inflated Pollen Counts'. Environmetrics, . [DOI] [Details]
M. Salter-Townshend, J. Haslett, N. Friel (2009) 'Discussion of `Approximate Bayesian inference for latent Gaussian models by using integrated nested Laplace approximations' by Rue, Martino and Chopin'. Journal of the Royal Statistical Society, Series B, 71 . [Details]
Salter-Townshend, M. and Murphy, T.B. (2013) 'Variational Bayesian inference for the latent position cluster model for network data'. Computational Statistics & Data Analysis, 57 (1):661-671. [DOI] [Details]
Salter-Townshend, M, White, A., Gollini, I. and Murphy, T.B. (2012) 'Review of Statistical Network Modelling: Models, Algorithms and Software'. Statistical Analysis and Data Mining, 5 (4):243-264. [DOI] [Details]
J. Haslett, M. Whiley, S. Bhattacharya, M. Salter-Townshend, S. Wilson, J.R.M Allen, B. Huntley, F.J.G. Mitchell; (2006) '\i Bayesian Palaeoclimate Reconstruction0'. Journal of the Royal Statistical Society Series A-Statistics in Society, 169 (3):1-36. [Details]
 

Conference Publications

Salter-Townshend, M. and Murphy, T.B.; (2011) Sentiment Analysis of Online Media Proceedings of the 58th Congress of the International Statistical Institute Dublin, Ireland, Available Online [Details]
Salter-Townshend, M. and Murphy, T.B.; (2012) Sentiment Analysis of Online Media . In: Lausen, B., van del Poel, D. and Ultsch, A eds. Joint Conference of the German Classification Society (GfKl) and the German Association for Pattern Recognition (DAGM) [Details]
Michael Salter-Townshend and Anthony Brew; (2010) A Latent Space Mapping for Link Prediction Networks Across Disciplines: Theory and Applications [Details]
Thomas Brendan Murphy and Michael Salter-Townshend; (2009) Variational Bayesian Inference for the Latent Position Cluster Model Analyzing Networks and Learning With Graphs; NIPS [Details]
J. Haslett, A. Parnell and M. Salter-Townshend; (2006) Modelling Temporal Uncertainty in Palaeoclimate Reconstructions International Workshop on Statistical Modelling [Details]
M. Salter-Townshend and J. Haslett; (2006) Zero-Inflation of Compositional Data International Workshop on Statistical Modelling [Details]
         

Dissertations/Theses

Michael Salter-Townshend; (2009) Fast Approximate Inverse Bayesian Inference in non-parametric Multivariate Regression (with application to palaeoclimate reconstruction). Dublin: Dissertations/Theses Available Online [Details]
                                   

Media

Salter-Townshend, M (2012) Analysing My Facebook Friends. Media Available Online [DOI] [Details]
                                       

Research

Research Interests

Statistical and probabilistic modelling of social networks:
 - Fast approximate inference routines for Bayesian latent variable network models.
 - Multiple-relations / multiview / multilevel network modelling.
 - Community finding and block modelling.
 - Link and covariate prediction.
 - Role analysis of nodes in a network via ego-network modelling.
Inference Methodology:
 - Variational Bayes.
 - MCMC.
 - Integrated Nested Laplace Approximations (INLA).
 - EM algorithm.
Sentiment Analysis:
 - Modelling multiple imperfect annotations.
 - Model Based Clustering of annotators.
Palaeoclimate Reconstruction:
 - Modelling multivariate, zero-inflated, constrained, spatial counts data.
 - Solution of Inverse Problems via inversion of Bayesian Hierarchical Models.
     

Collaborators

Internal Collaborators

Prof. Brendan Murphy

External Collaborators

Prof. John Haslett