Two shortcourses on Bayesian Modelling and Statistical Machine Learning.
June 1115, 2018
Venue: Room 5027 (5A) in the Mathematics Building (Number 54).
Course 1: Introduction to Bayesian Hierarchical Modelling and Computation, June 1112, 2018.
The first shortcourse is aimed at applied scientists (with good graduate degrees) who are thinking of using Bayesian methods and would like to receive a gentle introduction with a large practical component using R and WinBuGS.
No previous knowledge neither any rigorous training in mathematical statistics is required. No previous knowledge of Bayesian methods is necessary. However, familiarity with standard probability distributions (normal, binomial, Poisson, gamma) and standard statistical methods such as multiple regression will be assumed.
Theory lectures on the Bayes theorem, elements of Bayesian inference, choice of prior distributions and introduction to MCMC will be followed by handson experience using R and the WinBUGS software.
Course 2: Statistical Machine Learning. June 1315, 2018.
This course will provide an overview of basic ideas in statistical machine learning. The course should be useful to applied scientists from any discipline wo would like to use statistical machine learning in their research and data analysis. No previous knowledge neither any rigorous training in mathematical statistics is required.
The topics to be covered include: Supervised learning, classification, algorithms and unsupervised learning.
The course begins with detailed discussion of supervised learning. It will discuss the subtopics of regression where the usual topics of multicollinearity, variable selection, regularisation, LASSO prior, Ridge prior, Elastic Net prior
will be illustrated with many examples. This will be followed by the discussion of nonlinear regression where we will also consider the topics of Gaussian Process Prior Regression.
The topic of Classification will be discussed with special emphasis on the naive Bayes classifier, Discriminant Analysis, logistic regression, Decision Tree, Support Vector Machine, Random Forest, Perceptron Learning, Neural Network and Deep Learning.
Next, we will discuss various popular algorithms such as the Gradient Descent, Stochastic Gradient Descent and Back Propagation.
Unsupervised learning is the last major topic to be discussed in this course. Here we will introduce the Kmeans clustering, principal component analysis and latent Dirichlet analysis.
Both of the above courses will have a large practical hands on traning component for which participants are required to bring their own laptop. Methods will be illustrated using several practical examples from finance, system bilogy, social sciences etc.
Rcode and data sets will also be provided at the beginning of the courses.
Who should attend? The two courses are primarily aimed at applied scientists who wish to use Bayesian methods and statistical machine learning ideas in their data analysis and modelling problems. The courses will be suitable for applied scientists and statisticians from government departments, practitioners from industry, and research students at all levels.
Academic researchers and scientists from other disciplines can also attend but should have some background in statistics/mathematics to fully understand the whole course.
Prerequisite for Course 1 (Hierarchical Bayesian Modelling): Participants should have a graduate degree any numerate discipline with some experience in analysis and presentation of data. Participants must have some familiarity with standard probability distributions (normal, binomial, Poisson, gamma) and standard statistical methods such as multiple regression. No previous knowledge of Bayesian methods is necessary. Basic familiarity with the Rsoftware package is also desirable. Participants should bring their own laptop.
Prerequisite for Course 2 (Statistical Machine Learning): A graduate degree with some experience of handling large data sets is required to fully appreciate the materials to be presented in this course. Participation in the previous shortcourse on Bayesian methods is not compulsory but desirable.
Please email Professor Sahu (S.K.Sahu@soton.ac.uk) who can advise regarding the prerequisite.
Registration Information
Course 1: Hierarchical Bayesian Modelling, June 1112, 2018.
Research students  £300 
Academics  £400 
All others  £500 
Course 2: Statistical Machine Learning, June 1315, 2018.
Research students  £450 
Academics  £600 
All others  £750 
 The fee will include course materials, computing facilities, lunch and refreshments each day.
 University of Southampton staff and students will receive a 25% discount on the above costs.
 Please email the professional training secretary if you require any assistance.
Payments can be made by the University online store by clicking the links below.
 The number of spaces is limited, so an early registration is advised.
 Fees can be refunded in full if cancelled before May 11, 2018.
 Participants are required to book their own accommodation.
About the Lecturers
Prof Sujit Sahu (University of Southampton) is an expert in model based Bayesian data analysis and has 25 yearsâ€™ experience in this area of research. His research strength is in practical hierarchical Bayesian modelling and MCMC computation. He has successfully delivered similar shortcourses in Bayesian statistics in Australia, Chile, Italy and Spain and also biennially in Southampton since 2005.
Dr Sourish Das is an Assistant Professor in the Chennai Mathematical Institute, India. Sourish obtained his PhD degree from the University of Connectcut in the USA and his current research interest focuses on solving problems of Financial Mathematics, Statistical Machine Learning, and Bayesian methodology. He has published many articles on statistics and machine learning with applications to Big Data. He has delivered similar shortcourses on a number of occasions in the Indian multinational corporations.
