Curriculum (40 units)
Students must complete a total of 40 units for the programme, which includes 16 units from compulsory core courses and at least 16 units from the list of AIS electives. Additionally, students have the option to enrol in up to two courses (8 units) at the 4000-level from other departments within the Faculty of Science or opt for two courses (8 units) at the 5000-level offered by other departments, subject to approval. Workloads are counted in terms of course units, with one regular course worth 4 units. Academic performance is evaluated using grade points on a 5-point scale, with a minimum Grade Point Average (GPA) of 3.00 required for graduation.
Core Courses
Students must complete 4 core courses (total 20 units)
Machine learning (ML), in most scientific applications, is a computational model of statistics. This course will give students a broad overview of why and how ML is used in the sciences. It will also provide some context for how ML tools were developed to solve certain classes of challenges, highlighting the unique requirements of ML in science. For example, the use of ML in curiosity-driven exploration, understanding how the data was prepared and labelled, determining (where possible) whether the use of ML was ill-posed or ill-conditioned, interpreting the predictions of these ML models, and turning these into hypotheses.
This course exposes graduate students to the computational basics of Machine Learning and Data Science commonly used for exploration and discovery in the sciences (e.g., optics, statistical physics, condensed matter physics, structural biology, chemistry, materials science, and epidemiology). We will take a hands-on approach to building, implementing, training, and evaluating machine learning models (in Python), through examples of discovery and exploration in scientific applications.
This course presents the mathematical and computational foundations of machine learning, preparing the students with sufficient background for more advanced topics such as natural language processing. The learning outcomes include adequate familiarity with the programming environment for machine learning with Python, a deeper understanding of the building blocks of neural networks, and numerical training algorithms for machine learning. The course will draw science applications to illustrate deep learning concepts.
This course will be based on a series of seminars by leading AI practitioners in Singapore and overseas, both in academia and industry. The idea is to expose students to rapidly changing trends in AI and breakthroughs in applying ML to existing and future problems. Doing so will help students appreciate how tangible tools they learned in other courses in this programme can impact research. Students will think critically about and discuss the content of these seminars with their peers. Students will also write short summaries of these seminars that will be assessed.
Elective Courses
Astronomy and astrophysics are transforming in the era of Big Data, with massive datasets of celestial phenomena collected by next-generation telescopes. This course explores the emerging role of AI in accelerating and enhancing the scientific analysis of these astronomical datasets, facilitating discoveries, and addressing fundamental questions about our Universe. Students will gain insight into how AI optimizes and improves upon modern techniques utilized in astronomy, with key topics including the data mining of astrophysical measurements, time-domain analyses, spectral analysis techniques, and probabilistic inference. Through lectures and hands-on projects, students will explore observations of a variety of celestial objects, including exoplanets, variable stars, and galaxies. They will develop an understanding of the science underlying celestial phenomena and form a problem-solving mindset utilizing AI, focused on tackling intriguing problems in astronomy and astrophysics.
Imaging has become highly data-driven and reliant on computation. New imaging modalities have emerged that incorporate known principles of optics (e.g., beam shaping, propagation), beam-matter interactions, and priors about the sample. These have resulted in a new form of imaging known as computational lenses, which does far more than what physical lenses alone can accomplish. The development of computational lenses is accelerated by machine learning, affordable and fast computing, and high-throughput data collection. This course will equip students with the foundations needed to engage and extend imaging empowered by machine learning. We start by building up the essential optics foundations in imaging and practical aspects of detection physics. Then, students will learn about computational methods (both conventional and machine learning) used in analysing raw images, for instance, separation of signal from noise in detected images, computational phase retrieval, de-noising, computed tomography, segmentation, etc. Finally, we will explore how deep learning is changing this landscape. Throughout the course, students will see how these concepts and tools are being applied in imaging examples in biology.
Applications of AI in Science have been rapidly changing. Many of these applications are built on the foundations of machine learning and data science that are covered in the core courses of this programme. In this course, faculty member(s) from the Faculty of Science will present emerging applications of AI in science (e.g., condensed matter physics, quantum physics, optics, statistical mechanics, chemistry, biology, materials science, etc) in a pedagogical manner.
Students will get to experience solving a scientific research problem using machine learning. Students are expected to contact faculty members in the Faculty of Science as potential supervisors for this course. Both parties will seek to define the types of datasets and problems they might have that are amendable to machine learning. Students will complete a research project (normally within two semesters) under the supervision of academic staff, submit a project thesis, and present the results to the supervisor(s) and the examiner(s).
Students will spend a semester solving AI-centric problems with companies and organizations. Some of these companies are leaders in applied or theoretical AI. This experience will help students understand how theoretical tools are translated into practical outcomes. Liaisons will be identified within these companies, each of whom will propose specific projects that can be done by up to three students in a group.
Modelling and optimization have been the mainstay in both AI and computational physics. This course covers computational techniques for solving physics and engineering problems, emphasizing molecular simulation and modelling. Topics will be from the text, “Numerical Recipes” by Bill Press et al supplemented with examples in materials and condensed matter physics. This course ensures that graduate students intending to do research in computational physics will have sufficient background in computational methods and programming experience.
Quantum computing is expected to fundamentally change our relationship with AI. This course provides an intro-duction to quantum information and quantum computation. In addition to physics majors, the course addresses students with a good background in discrete mathematics or computer science. The following topics will be covered:
(1) Introduction: a brief review of basic notions of information science (Shannon entropy, channel capacity) and of basic quantum kinematics with emphasis on the description of multi-qubit systems and their discrete dynamics.
(2) Quantum information: Entanglement and its numerical measures, separability of multi-partite states, quantum channels, standard protocols for quantum cryptography and entanglement purification, and physical implementations.
(3) Quantum computation: single-qubit gates, two-qubit gates and their physical realization in optical networks, ion traps, quantum dots, Universality theorem, quantum networks and their design, simple quantum algorithms (Jozsa-Deutsch decision algorithm, Grover search algorithm, Shor factorization algorithm).
This course is tightly integrated with IBM quantum computer hands-on experience via IBM Q Experience cloud services. Students will learn the fundamentals of Qiskit, a modern and rapidly developing quantum computer programming language, by directly implementing concepts learned in the classroom.
Machine learning relies on data, and real data has to be interpreted statistically. In the age of big scientific data, Bayesian statistical methods and machine learning techniques are becoming a vital part of the modern scientist’s toolkit. This course provides a graduate-level introduction to the two related fields, emphasizing both equally. Key topics for the first part include: fundamentals of probability and inference, hierarchical modelling, model validation and comparison, and Monte Carlo methods; for the second part, they include: classification and regression, kernel methods, variational methods, and neural networks. This course will be largely theoretically oriented, with the occasional computational component.
AI is poised to help us tackle complex problems and model complex systems. Much of our real-world data are manifestations or measurements of their underlying complex interactions. Hence, modelling and analysing the underlying complex systems can reveal understandings and predictions that complement black-box machine learning tools. This course will cover the basic concepts and tools for analysing complex systems and simulation models and, more importantly, why and when we need such white-box tools derived from statistical physics. Certain fundamental concepts in complexity science will be introduced. It will also provide hands-on experience with system analysis and simulation modelling in Python.
In a growingly digitalized world, chemistry continues to be transformed by the availability of big data sets and powerful artificial intelligence algorithms. This course offers a broad-based learning opportunity for students to explore a plethora of data analysis and machine learning techniques in the chemistry perspective. Specifically, technical topics that will be covered include dataset transformation, basic statistical analysis, feature selection, popular learning algorithms and generative modelling. The application and relevance of these techniques will be exemplified by up-to-date chemical examples, such as, synthesis planning, inverse design of functional molecules, polymer engineering and small molecule drug design.
This course is designed to provide students with a comprehensive understanding of deep learning, a pivotal machine learning technique with extensive applications in artificial intelligence and data sciences. The curriculum focuses on the introduction of fundamental concepts, numerical algorithms, and computing frameworks pertinent to deep learning. Special emphasis is placed on numerical algorithms and their implementation within industrial computing frameworks, as well as the analysis of data-intensive problems derived from real-world applications. Core topics covered in this course include: Foundations of Neural Networks; Optimization Strategies in Deep Learning; Representative Network Architectures; Regularization and Generalization; Model Evaluation and Tuning; Deep Learning Frameworks and Implementation Applications in Natural Language Processing and Computer Vision.
This course equips students with foundational knowledge and skills for textual data processing and analysis. It covers essential text processing and related machine-learning algorithms. The goal is to enable mastery of text processing techniques, understand core principles of natural language processing, and gain experience with applying machine learning tools for text processing and interpretation. Major topics include the basics of text processing, foundational principles of natural language processing, essential machine learning algorithms for text analytics, introduction to deep learning and its applications in text processing and interpretation.
This course provides a comprehensive exploration of Natural Language Processing (NLP) techniques, with a focus on their practical applications in data science. Students will gain a strong foundation in text preprocessing, tokenization, and text analytics, and progress to advanced topics such as named entity recognition, text classification, and sequence to sequence modelling. The course also delves into Large Language Models (LLMs), covering their architecture, training methodologies, and real-world applications. Students will learn to develop NLP applications and harness LLMs to automate tasks, generate content, and tackle complex language-related challenges. This course equips students with the skills needed to excel in data science and leverage NLP in various industries.
This course dives into how Artificial Intelligence [AI] is transforming the pharmaceutical industry in various segments, focusing on targeted, personalized, and adaptive therapies. Via seminars by faculty members and invited experts, students in this course will learn how AI enhances drug delivery technologies through sophisticated data analysis, pattern recognition, and optimization. Students will also explore about how computational pharmaceutics and big data are applied in the optimization of drug formulations and improving manufacturing processes. This course prepares students for the future of drug product development with the knowledge of AI and big data, where the industry is moving fast to era 5.0.
Cross-faculty/Department courses
Students may take up to two science elective courses 4000 or 5000 from any department in the Faculty of Science, conditional on the students being able to enrol in these courses. The registration for any cross-faculty/department course is subject to approval by both the programme and the department that offers the course.