Presentation of the paper:
"Context-adaptive Modelling and Inference"
 
Speaker: Norman Poh

Abstract: 
Real data samples are affected by context. For instance, in the medical domain, age, gender and ethnic background often constitute the context. The variables that represent the context are used in Economics, Psychology, Statistics and many other domains. These variables are known as co-variates, contextual variables, control variables, exogenous variables and exploratory factors. In this talk, we shall examine a data-driven approach that attempts to model the context and use it for inference (prediction). A few issues such as context extraction, context modelling and context transfer will be explored. Preliminary case studies will be illustrated using biometric person recognition and modelling of estimated glomerular filtration rate (eGFR), a measurement that is often used for diabetic patients.