Domain Adaptation using Passive and Active Approaches
Presenter: Dr Haider Raza
Affiliation: Institute for Analytics and Data Science, CSEE, University of Essex
Systems based on machine learning methods often suffer a major challenge when applied to the real-world datasets. The conditions under which the system was developed will differ from those in which we use the system. Few sophisticated examples could be email spam filtering, stock prediction, health diagnostic, and brain-computer interface (BCI) systems, that took a few years to develop. Will this system be usable, or will it need to be adapted because the distribution has changed since the system was first built? Apparently, any form of real-world data analysis is cursed with such problems, which arise for reasons varying from the sample selection bias or operating in non-stationary environments.
Theoretical and Practical Part:
This tutorial will focus on the issues of domain adaptation under passive and active approaches for dataset shifts (e.g. covariate shift, prior-probability shift, and concept shift). We will discuss only domain adaptation under the issue of covariate shift. Since working on unlabelled data is widely available across all domains of application. Moreover, we will cover the theoretical definitions and examples of both prior-probability shifts, and concept shift. We will also see the methods to identify covariate shift and the proper measures that can be taken in order to improve the predictions.