Graduate Student Tutorials

NUS School of Computing - Graduate Student Tutorials 2018

The two-day tutorial sessions organised by NUS School of Computing provides an excellent opportunity for student participants to explore as well as to gain exposure and foundations to the latest computing research. This forum, which is particularly useful for students who wish to embark on computing research, also aims to facilitate interactions and discussions of ideas amongst student participants, researchers and faculty members.

Graduate students and final-year undergraduate students are welcome to attend the tutorials. Participants from external institutions or organizations may also apply to attend the sessions. 

Date & Time
6 & 7 August 2018
9.00am to 5.00pm

Breaks: Morning – 10.30am to 11.00am
             Lunch – 1.00pm to 2.00pm
             Afternoon – 3.30pm to 4.00pm

NUS School of Computing (COM1 Level 2)

Registration Period        : 4 to 24 July 2018
Registration Fees           : S$200 per day (Industry participants only)
Registration Link            : Closed

Due to limited places, attendance is strictly by registration only. Only notified registered participants will be allowed to attend the sessions.

Programme Outline

DAY 1 – 6 August 2018
Session CS-1
Title Hands-on Optimization of CPU and FPGA Designs with a Modern Computer Architecture Overview
Speaker Trevor E.Carlson Assistant Professor, Dept of Computer Science
Venue Video Conference Room (COM1-02-13)
Abstract While computers continue to advance in performance today, the near future will bring challenges for computing. We can no longer continue to add additional transistors to our processors as the costs have begun to rise (Moore’s Law is ending). We can no longer turn on more transistors at the same time because our processors are getting too hot (Dennard Scaling is ending). To address these problems, computer architects have turned their focus to efficiency, a departure from previous years’ efforts of performance at all costs. The goal of this course is to provide a preview of some of the state-of-the art techniques developed today to address energy and area efficiency challenges in computer architecture. We will focus on developments that apply to today’s devices, from watches to smartphones, as well as desktops and servers. More specifically, this class will begin from the transition to multi-core technologies, discuss energy-efficient techniques that allow these general-purpose architectures to improve, and finally, describe the latest techniques and architectures, like the focus on memory-level parallelism (MLP), load-slice processors and out-of-order commit processors as well as FPGA accelerators. In the second half of this course, we will conduct a hands-on coding and optimization playground. We will take a deep-dive into the latest software computer architects use to design next generation processors and accelerators. In small groups, participants will choose from a number of platforms and tutorials to learn how to become the computer architects of the future. Students can focus on one or two platforms, and work in-depth to improve state-of-the-art designs, or take a broader view, and work through the basic tutorials for each platform to understand how each platform contributes in different ways. Example platforms include optimization of Machine Learning algorithms on personal Xilinx PYNQ FPGAs, and experimenting with the Sniper Multicore Simulator to evaluate next-generation IoT and server applications. At the end of the second half, each student group will give a short presentation on one of the topics they found interesting during the hands-on session, and explain its connection to computer architecture and the future of computing.
Session CS-2
Title Gaussian Process Methods in Machine Learning
Speaker Jonathan Scarlett Assistant Professor, Dept of Computer Science
Venue Seminar Room 3 (COM1-02-12)
Abstract I will provide an introduction to a powerful tool in modern machine learning and statistics known as Gaussian Processes (GPs). This is a rich and versatile class of probabilistic models for modelling input-output relationships in data, with recent successful applications including tuning deep neural networks, optimising robot parameters, modelling environmental data, and more. In contrast with parametric methods that assume a very specific type of underlying relationship (e.g., a linear function), GP methods can directly model more complex relationships in a principled manner. I will introduce GPs in the context of regression, i.e., learning an entire input-output relationship of the form y = f(x) based on samples of (x,y) pairs. I will then move onto more recent developments in Bayesian optimisation, where one adaptively samples in order to search for an input maximising the underlying function f(x).  I will also discuss some of the main ongoing challenges in the field, including scaling to large data sets, and handling “high-dimensional” scenarios with a very large number of input variables.
Session CS-3
Title Biometrics Authentication
Speaker Terence Sim Associate Professor, Dept of Computer Science
Venue Seminar Room 9 (COM1-02-09)
Abstract Biometrics is increasingly being used for authentication, eg. fingerprint for phone payment, face recognition for door access. But how does biometrics work? What are the challenges? This one-day course will give an overview of the different types of biometrics, examine their pros and cons, and introduce the concepts of pattern recognition that underlie all biometrics authentications. A hands-on session will be conducted for you to experiment with pattern recognition.
Session IS-1
Title Understanding the Dynamics of a Digital Platform Ecosystem: Application of Network Analysis and Deep Learning
Speaker Um Sungyong Assistant Professor, Dept of Information Systems and Analytics
Venue Seminar Room 10 (COM1-02-10)
Abstract Digital platform ecosystems are dynamic: they grow and evolve as new firms join the ecosystems. However, the way digital platform ecosystems change is not clearly understood. Qualitatively, this course introduces the two basic conceptual approaches to understand the structure of a digital platform ecosystem. Quantitatively, this course draws on computational statistical methods to capture the structural change of a digital platform ecosystem. First, from the perspective of deep learning, this course introduces the way that deep learning techniques such as Word2Vec are used to sort out network information from various resources such as firm documents and application source codes. Second, from the perspective of network analysis, this course focuses on interactions of which form a complex bipartite network that drive the changes in the topological structure of a digital ecosystem over time. In particular, changing combinations of existing digital components such as APIs and software packages in applications are focused on to capture the dynamics of a digital platform ecosystem.
DAY 2 – 7 August 2018
Session CS-4
Title Security Challenges and Solutions of the Internet and its Things 
Speaker Kang Min Suk and Han Jun Assistant Professors, Dept of Computer Science
Venue Video Conference Room (COM1-02-13)
Abstract The Internet has evolved at a fast pace in the past few years and embraced numerous emerging networks, including Internet of Things, 5G, and blockchain networks. At this exciting moment of rapid transformation, unfortunately, we have faced new security challenges in the newly expanded attack surface. In this lecture, we will discuss security challenges in (1) designing highly available next-generation Internet, (2) secure operation of large numbers of Internet of Things devices, (3) fixing vulnerabilities in 5G LTE systems, and (4) designing secure and private blockchains. Our work has addressed several fundamental questions of these security problems, often utilizing new security primitives. The lecture may have some hands-on sessions, and thus students are encouraged to bring their laptops to enjoy the demonstrations.
   Session CS-5
Title Privacy in the Age of Machine Learning
Speaker Reza Shokri Assistant Professor, Dept of Computer Science
Venue Seminar Room 3 (COM1-02-12)
Abstract What does digital privacy mean? What is anonymity? How can a computer scientist reason about privacy in a quantitative manner? What can a curious entity learn about our sensitive information in a computing system? Is it possible to share our data without risking our privacy? What are the risks of releasing aggregate statistics or machine learning models? How can we protect data privacy? What is the role of encryption and randomization?   I will provide an algorithmic framework for answering these questions for real applications. I will talk about how to assess privacy risks and to guarantee privacy in various computing systems including machine learning. Basic probability and algorithms are the pre-requisites for this course.
Session CS-6
Title Analytical Performance Modeling for Computer Systems
Speaker Tay Yong Chiang Professor, Dept of Computer Science
Venue Seminar Room 9 (COM1-02-09)
Abstract In predicting and analyzing the performance of a computer system (be it hardware or networking or databases, etc.), one powerful tool lies in mathematical modeling. This course will introduce some basic techniques (Little’s Law, bottleneck analysis, decomposition, etc.) and illustrate them with papers from the literature. It is based on the book “Analytical Performance Modeling for Computer Systems” (Morgan & Claypool).
Session IS-2
Title Digital Innovation: Business Models and Computational Analysis
Speaker Hahn Jungpil and Jack Jiang Associate Professors, Dept of Information Systems and Analytics
Venue Seminar Room 10 (COM1-02-10)
Abstract This course teaches students how to create innovation-driven business model through both process innovation and product innovation. Students will learn how to identify opportunities for technology innovation, particularly disruptive innovation, and how manage innovation process. Cases will be used to facilitate students to understand the value of digital innovation and the importance of business ecosystems.  The second half of the course will focus on methodological approaches to researching digital innovation with focus on computational modeling and agent-based simulation approaches.