Deep Reinforcement Learning

This course will introduce both basics and advance d topics of deep reinforcement learning. Specific topics include machine learning, deep learning, reasoning and planning under uncertainty, (deep) reinforcement learning, and graphical models. The main goal of the course is to equip students with the tools to tackle new AI problems you might encounter in life and also to serve as the foundation for further study in deep reinforcement learning.

Security Technologies in Cyberspace

Cyberspace security consists of three parts – information security, system security, and application security. Information security leverages cryptographical mechanisms to guarantee data privacy; system security leverages sandboxing, program analysis etc. to guarantee the authentication and privacy of system usage; application security leverage anomaly behavior detection to recognize malicious user and software. In this course, we would discuss typical security issues and solutions in each section, and require students to practice them in course projects. We would also research into new security issues in latest scenarios and new technologies (e.g., machine learning, trusted execution environment) that can be used to solve security problems.

Algorithm Economics

This course develops a framework to analyze the algorithmic-service industry organization and regulation by an interdisciplinary perspective of economics and computer sciences. Algorithm design has deeply involved into the operation of the platform economy, the dispatch economy, and the sharing-and-renting economy. The algorithm designs shape the market equilibriums, decide the welfare-distribution, and affect the environments and global climate change, Symmetrically, the future human society also eagerly expects the research about how government can develop “smart regulations” for the algorithm economy. All these topics are the frontier of the interdisciplinary research of economics and computer science. Thus, this course will include two sections. The first section discusses the algorithm design in the modern economy, such as the algorithm designs and associated economics theories for the platform economy, dispatch economy, and sharing-and-renting economy. In particular, we will introduce the dispatch algorithm designs of the electricity economy; the bidding and matching algorithm of the sharing-and-renting economy. The second section focuses on the economic impacts of the algorithm design, including the influence of the algorithm design on the market equilibrium, algorithm design and market failure, algorithm design and welfare distribution. In particular, we will discuss the environmental and economic impacts of the dispatch-algorithm design and prediction-algorithm development in an electricity market. The model of private-information market. Price discrimination and bidding algorithm of the sharing-and-renting economy. The course will embed the algorithm models into the economics framework. The knowledge teaching will be coupled and demonstrated by the literature reading and reviewing.

Stochastic Network Optimization Theory

This course will give an in-depth introduction to the recently developed Lyapunov optimization theory for stochastic networks. It aims at introducing to the students various concepts of queue stability, general models for stochastic queueing networks, the minimum-drift algorithm design principle, and the Lyapunov drift analysis technique. It will also present applications of the theory to both networking and operations research problems, and encourage the students to apply the theory to their own problems of interest.

Selected Topics in Information Physics

This course aims at introducing graduate students to the rapidly developing field of information physics. Thecourse will highlight how the most important concepts in computation and complexity theory are physical, and howsome of the most puzzling question in physics can be answered through information theory. We review classicalinformation theory, and how to must be generalized with the discovery of quantum mechanics. The ramifications ofthis to information storage, energy extraction and computational complexity are discussed.

Quantum Electronics and Advanced Atomic Physics

This course provides a practical knowledge of quantum electronics and advanced atomic physics for graduate students who are performing atomic and optical experiments. First, we provide a fairly conventional discussion of Gaussian beams, cavities, nonlinear optics and modulation techniques. Then we seriously discuss the knowledge of atomic structure and atom-photon interaction. Finally we connect them for the amplification of light and spectroscopy for the laser frequency stabilization. A number of very recent developments are discussed, such as frequency metrology using femtosecond lasers, laser cooling and trapping, and Ion traps.

Advanced Quantum Information Theory

This course provides an overview of the latest advancements in quantum information theory and quantum computation. Starting from the foundation of the framework of density matrices and open system dynamics, the course gradually transitions into two main chapters, the first about the quantum theory of information compression and transmission and the second on the topological models of quantum computation. The first chapter will deal with quantum entropies and quantum typicality, providing useful tools also for students interested in quantum thermodynamics. The second chapter will deal with anyons and geometric phases, showing the features of anyonic statistics can be used to perform stable and efficient quantum computations.Students who take this course will develop the mental discipline needed to identify and discuss critically these questions and will be provided with the sharpest theoretical tools to address these questions.

Quantitative Financial Credit and Risk Models

Internet brings significant impacts on finance industry, starting with the credit models. On the positive side, new technology enables new markets that potentially benefit more people in the society. However, the technology calls for a complete new credit model, which is still a state-of-the art research problem. Even worse, the fraudulent activities on the Internet bring more challenges to the finance industry. In this course, we present recent academic and industry practice in this area, to get students started on this cutting-edge research direction. There are three modules: 1) Credit models and architecture; 2) Anti-fraurd architecture and design; and 3) Case studies. The students are expected to read 2 recent publications per week on average, and complete two significant course projects. On project is based on an open dataset and the other is open-ended.

Hot Topics in Computational Biology

The course covers research progress and hot topics in Computational Biology and introduces topics including basic computational theory and methods, three-dimensional structure determination and dynamic study of proteins, protein and drug molecular design, Proteomics, and Biology evolution model.

Advanced Computational Economics

The course covers classic and state-of-the-art results on computational and game-theoretic questions related to computational economics.

Introduction to Computational Theory

This course is set for junior graduate students, who are expected to have good understandings of mathematics and knowledge in basic theoretical computer science.

The course introduces main contents of computational theory with an emphasis on important topics concerning modern and contemporary complexity theories, which enables students to know major issues and results in the field of complexity theories. It also helps students to determine their future research interests through understandings of algorithm.

Topics covered in the course are: review of main contents of computability theory, introduction of basic themes of complexity theory, including basic complexity classes like P, NP, PSPACE and BPP; proof of the non-existence of circuit and Parity in AC0; Hierarchy Theorems of time and space; main results of derandomization; PCP theorem and non-approximatability; theoretical coding; and etc.

The course is mainly conducted through lectures and series seminars, supplemented by featured discussions. The students are required to take thesis reading exercises and give summary reports, with a view to helping them find their future research interests.

Algorithm Analysis and Design

This course is set for junior graduate students, who are expected to have good understandings of mathematics and knowledge in basic theoretical computer science.

The course introduces advanced technologies concerning the design and analysis of algorithm, as students will read theses in the field of algorithm design. It also helps students to determine their future research interests through understandings of algorithm.

Topics covered in this course are: review of basic technologies of algorithm design including divide-and-conquer algorithms and dynamic programming; introduction of the design and analysis of random algorithm and approximation algorithm; introduction of the current research on important issues including linear algorithm, online algorithm, and data structure in computational geometry.

The course is mainly conducted through lectures and series seminars, supplemented by featured discussions. The students are required to take thesis reading exercises and give summary reports, with a view to helping them find their future research interests.

Advanced theoretical computer science

This course is set for junior graduate students who are interested in theoretical computer science and relevant disciplines. Those registered for this course are expected to have sound foundation Discrete Math, Algorithm Design, basic machine learning knowledge.

Lectures for this course, all in English, aim to introduce current research directions, latest development and hot topics in computer science, and carry out in-depth discussions on issues of common interest. This course will help students to determine their future research interests and goals through featured discussions.

The course quickly recall basics about convex optimization and machine learning: linear/logistic regression, regularization, newton method, stochastic gradient descent (asychronized, variance reduction method), generative vs discrimitive, variance vs bias. Off-the-shelf machine learning and prediction algorithms: k-nn, SVM, kernel trick, clustering, Adaboost, gradient boosting, random forest. Online learning and sequential prediction. Multi-armed bandit, Universal portfolio, Multiplicative weighting method, online convex optimization, basic time series. linear algebra-based learning algorithms: SVD, principle component analysis (PCA), independent component analysis (ICA), Nonnegative matrix factorization (NMF), topic modeling, matrix completion, dictionary learning, tensor method, spectral clustering.

The course is mainly conducted through lectures and series seminars, supplemented by featured discussions. The students are required to take thesis reading exercises and give summary reports, with a view to helping them find their future research interests.