Personalized Machine Learning

Syllabus

Recent advances in machine learning have enabled a number of applications for health and well-being, marketing and social robots, among others. Traditional machine learning relies mainly on generic models: models tuned to an average target population. However, the ‘good’ performance by these generic models doesn’t necessarily translate to each individual in the group. While this can be acceptable in certain domains (e.g., marketing research), when it comes to, for instance, health and well-being, new systems need be optimized and work for each person. They should also help an individual to see, for example, which factors they might change in their life to improve their health or mood. Likewise, in order to build an adaptive robot or learning system, we might want to learn which factors are influencing each learner's engagement the most.

Context is also a key element -- how do we learn the influences of context and social environment on each participant's performance and embed that information in personalized models? This also brings a number of modeling challenges: (i) how can the contextual information (e.g., who the person is, what is his/her task, and so on) be combined efficiently and effectively to achieve better human-machine intelligence? (ii) How different (multi-modal) sensory data (e.g., video, audio, physiological, text and mobile) can be combined to obtain more reliable and robust personalized machine learning models? (iii) Last, but not least, how can we include, in an interactive fashion, the target person in the learning/prediction ‘loop’ to make his/her feedback and/or personal goals an integral part of the personalized models?

The goal of this course is to address these domains, and provide the students with the machine learning approaches and tools that can be used to build personalized machine learning models for their application. Specifically, the course will examine the state-of-the-art machine learning approaches, including active learning, domain adaptation, deep neural networks and Gaussian Processes. The emphasis will be on human data analysis for health and wellbeing (e.g., prediction of mood, depression, pain and engagement). Other projects related to personalized machine learning (e.g., with applications to human-computer interfaces & human-robot interaction) are also welcome. Students who take the class for credit will be required to present a related paper and lead discussion. In particular, each student (or a group) will design a project where they will evaluate the effects/benefits of the personalized machine learning. This will include the selection, adaptation and evaluation of (one or more of) the above-mentioned machine learning approaches for their project (together with the course instructors).