Description

Machine learning and data analysis have revolutionized many scientific fields and industrial sectors. Much of this success hinges on the vast amount of digital data generated and collected in today's world. However, as we navigate this data-driven frontier, we face novel challenges emerging from the scale and societal implications of data usage.

In this seminar, our focus is on studying new algorithmic advancements developed for learning algorithms within the context of these challenges. We will explore computational models with limitations in memory or time, which are more compatible with practical settings, and investigate the algorithmic solutions to fundamental learning problems in such models. An intriguing question we'll address is whether increasing the volume of data can compensate for memory or time constraints.

Another major topic in our seminar is the societal considerations of individuals' data usage such as privacy and fairness. We will investigate new technical methods to design efficient algorithms while preserving the privacy of individuals who participated in the data, with a particular focus on differential privacy. Additionally, we will examine recent advancements in algorithmic fairness.

Throughout this seminar, we will consider fundamental problems in learning theory and study them within the context of these challenges. Our primary goal is to acquaint ourselves with state-of-the-art findings in the field while gaining proficiency in foundational technical tools applicable to theoretical research endeavors.

Schedule