MIDS Electives

This course examines various computer engineering methods commonly performed in developing machine learning and deep neural network models. The focus of the course is on how to improve the training and inference performance in terms of model accuracy, size, runtime, etc. Techniques that are widely investigated and adopted in industrial companies and academic communities will be discussed and practiced. Programming practices on these techniques are designed with heavy utilization of the PyTorch package. Prerequisites: Computer Science 201 or ECE 551D or ECE 751D.

Principles of data analysis and modern statistical modeling. Exploratory data analysis. Introduction to Bayesian inference, prior and posterior distributions, hierarchical models, model checking and selection, missing data, introduction to stochastic simulation by Markov chain Monte Carlo using a higher level statistical language such as R or Matlab. Applications drawn from various disciplines. Not open to undergraduate students or students who have taken Statistical Science 360. Recommended prerequisite: Statistical Science 611 or the following: Statistical Science 210 and (Statistical Science 230 or 240L) and (Mathematics 202, 202D, 212, or 222) and (Mathematics 216, 218, or 221, any of which may be taken concurrently).

Provides an introduction to the machine learning technique called deep learning or deep neural networks. A focus will be the mathematical formulations of deep networks and an explanation of how these networks can be structured and ‘learned’ from big data. Discussion section covers practical applications, programming, and modern implementation practices. Example code and assignments will be given in Python with heavy utilization of PyTorch (or Tensorflow) package. The course and a project will cover various applications including image classification, text analysis, object detection, etc. Prerequisite: ECE 580, ECE 681, ECE 682D, Statistical Science 561D, or Computer Science 571D.

This course will extend the foundation laid in software tools for data science to allow for efficient computing involving very large data sets. This course will explore the use appropriate algorithms and data structures for intensive computations, improving computational performance by use of native code compilation, use of parallel computing to accelerate intensive computations, use appropriate algorithms and data structures for massive data set, and use of distributed computing to process massive data sets. Prerequisite: BIOSTAT 821 or permission of the Director of Graduate Studies. Credits: 2

This course ties the mathematical theory of algorithms and graphs to their practical implementations. Students will learn about the mathematical structures that for the foundations for the behavior and analysis of algorithms from a variety of domains, with a particular emphasis on graphs. Students will also tie that theory to practice by writing code to implement those algorithms, and comparing experimentally observed runtimes to those projected by the mathematical theory.

Designed to be a sequel to Statistical Science 711. The basic five topics are: martingales, Markov chains from an advanced viewpoint, ergodic theory, Brownian motion and its applications to random walks, Donsker’s theorem and the law of the iterated logarithm, and multidimensional Brownian motion, connection to PDE’s. For those who have not had 711, we will prove the law of large numbers using martingales and obtain versions of the central limit theorem from Donsker’s theorem. Course requires a knowledge of measure theory. Prerequisite: Statistical Science 711 or Mathematics 631.

Statistical modeling and machine learning involving large data sets and challenging computation. Data pipelines and data bases, big data tools, sequential algorithms and subsampling methods for massive data sets, efficient programming for multi-core and cluster machines, including topics drawn from GPU programming, cloud computing, Map/Reduce and general tools of distributed computing environments. Intense use of statistical and data manipulation software will be required. Data from areas such as astronomy, genomics, finance, social media, networks, neuroscience.

A data scientist needs to master several different tools to obtain, process, analyze, visualize and interpret large biomedical data sets such as electronic health records, medical images, and genomic sequences. It is also critical that the data scientist masters the best practices associated with using these tools, so that the results are robust and reproducible. The course covers foundational tools that will allow students to assemble a data science toolkit, including the Unix shell, text editors, regular expressions, relational and NoSQL databases, and the Python programming language for data munging, visualization and machine learning. Best practices that students will learn include the Findable, Accessible, Interoperable and Reusable (FAIR) practices for data stewardship, as well as reproducible analysis with literate programming, version control and containerization.
Prerequisite: Permission of the director of graduate studies

Students learn to program in C and C++ with coverage of data structures (linked lists, binary trees, hash tables, graphs), Abstract Data Types (Stacks, Queues, Maps, Sets), and algorithms (sorting, graph search, minimal spanning tree). Efficiency of these structures and algorithms is compared via Big-O analysis. Brief coverage of concurrent (multi-threaded) programming. Emphasis is placed on defensive coding, and use of standard UNIX development tools in preparation for students’ entry into real world software development jobs.