NUS – Institute of Data Science

As part of IDS’ efforts to encourage the use of data science in research to the broader university community, we are organizing a series of PhD-Teach-PhD workshops on selected data science topics of interest. The workshops will be led by IDS’ PhD students to acquaint PhD students from other disciplines on the current topics of interests in data science.

There were 4 workshops in total, These workshops were open to current NUS graduate students only and were conducted in Zoom meetings. Following are the workshops details and also recordings of the workshops for those who are interested in the topics.

Workshop 1: Uncertainty Quantification for Reliable Machine Learning

Presenters: Xiong Miao

Date: 19 September 2022

Abstract: In recent years, machine learning systems have become increasingly pervasive in numerous application domains and even our daily lives. In many domains, particularly those of a high-stakes nature (e.g., healthcare, law, and others), it is insufficient for a machine learning system to merely state its best prediction. Instead, there is substantial motivation to create systems that can explicitly quantify their degree of certainty or uncertainty, so that the associated predictions can be used more reliably while taking additional caution as needed. In this workshop, we will introduce some fundamental concepts such as model uncertainty and data uncertainty, introduce some common algorithms for estimating uncertainty, and discuss how uncertainty is used to solve problems such as calibration, misclassification prediction, and out-of-distribution detection. This workshop is aimed at PhD students familiar with at least the basics of machine learning.

Recording:

Workshop 2: Extraction of Causal Relations in Textg

Presenters: Fiona Tan An Ting

Date: 20 September 2022

Abstract: Causality describes a relationship between two entities or events, one of which is an Effect resulting from a Cause. Causality is a core cognitive ability of humans and important for many scientific fields. Causal Relation Extraction (CRE) aims to extract relations that exhibit causal meaning. Some tasks in CRE include identifying if a causal relation exists in the sentence, or at the finer-grained level, which words refer to the Cause and Effect event. These are challenging tasks since they tackle a deeper level of machine understanding, closer to human’s abilities. CRE is also important because it has many applications, like for summarization, concept extraction, causal knowledge identification for downstream NLP tasks (e.g. Question Answering) and for causal reasoning in NLP. In this workshop, we will provide a brief introduction to the tasks, datasets, and models of CRE, assuming only general machine learning background and ideally some prior exposure to NLP.

Recording:

Workshop 3: Introduction to Deep Learning Theory

Presenters: Liu Fusheng

Date : 21 September 2022

Abstract: Deep learning is an important subfield of machine learning that has witnessed tremendous advances in many areas, such as computer vision, speech recognition, machine translation, and game-playing programs. While the mathematical theory of deep learning has mostly lagged behind the practical advances, the field has given rise to a number of interesting perspectives distinct from classical learning theory. In this workshop, we will give a gentle introduction to some of these new theoretical ideas, and leverage some simple examples. Broadly, we will focus on three main aspects of deep learning: approximation, optimization, and generalization. This workshop is aimed at PhD students familiar with at least the basics of machine learning and the associated mathematical tools used (e.g., probability, optimisation, etc.).

Recording:

Workshop 4: Collaborative Machine Learning and Model Markets

Presenters: Wang Naibo

Date: 22 September 2022

Abstract: One of our most important faculties is our ability to listen to, and follow, one speaker in the presence of others. This is such a common experience that we may take it for granted; we may call it ‘the cocktail party problem.’ Since this problem was formulated, the quest for equipping machines with such an ability has never stopped. With the advent of deep learning, speech separation and speaker extraction represent methods for solving the cocktail party problems through engineering approaches. These techniques have facilitated many speech processing tasks in multi-talker acoustic environments including automatic speech recognition, speaker verification and hearing aid development. In this workshop, we will provide a brief introduction to the cocktail party problem and recent deep learning based techniques for solving it.

Recording: