Deep Learning is revolutionizing artificial intelligence tasks such as language understanding, speech and image recognition, machine translation, autonomous driving, etc. This transformative impact of deep learning, which tries to model the neural networks in brains, was recognized with Nobel Prizes in 2024 and Turing Award in 2018. This course provides a comprehensive introduction to deep neural networks with a focus on underlying principles and engineering applications. Students will explore the fundamental concepts, optimization techniques, and software tools of deep learning starting from the basics of perceptron and progressing to advanced neural network models with convolutions and attentions. The course emphasizes an engineering perspective, hands-on learning, and integrating theory with practice, The course also introduces latest methods to enhance the efficiency of training and inference in deep learning models and systems. Designed for students from diverse engineering disciplines, this course aims to equip them with the skills and knowledge to effectively apply deep learning in their respective fields.

Enrollment Requirements

Machine learning deals with the automated classification, identification, and/or characterizations of an unknown system and its parameters. There is an overwhelming number of application-driven fields that can benefit from machine learning techniques. This course will introduce you to machine learning and develop core principles that allow you to determine which algorithm to use, or design a novel approach to solving to engineering task at hand. This course will also use software technology to supplement the theory learned in the class with applications using real-world data.

Enrollment Requirements

Course Texts

Introduction to Machine Learning, 3rd Ed., Ethem Alpaydin, The MIT Press, 2014. (ISBN: 978-0262028189)

Schedule

This course covers the fundamental techniques for the design, analysis and layout of digital CMOS circuits and systems. Major topics include: MOSFET basics (structure and behavior of a MOSFET, CMOS fabrication, and design rules), detailed analysis of the CMOS inverter (static behavior, ratioed vs. ratioless design, noise margins, computing rise and fall times, delay models, resistance and capacitance estimation, design and layout of static CMOS logic gates, dynamic CMOS logic design, sequential circuit design (static and dynamic sequential circuit elements, clocking schemes and clock optimization), CMOS data path design. Graduate-level requirements include additional homework and term projects.

In this class, we investigate the state-of-the-art in reconfigurable computing both from a hardware and software perspective; understand both how to architect reconfigurable systems and how to apply them to solving challenging computational problems. The purpose of this course is to prepare students for engaging in research on reconfigurable computing. Initially, we review in detail the basic building blocks of most reconfigurable computers. Characteristics of FPGA architecture such as the organization of device logic and interconnection resources are examined to quantify hardware limitations. These physical limitations are then contrasted with computer-aided design issues such as the selection of circuit component locations in devices (the placement problem) and subsequent circuit interconnection between components (the routing problem). We then focus on the architecture for existing multi-FPGA systems and on compilation techniques for mapping applications described in a hardware description language to reconfigurable hardware. We will explore the question of “What makes an application suitable for reconfigurable computing?” with case studies in bioinformatics, image processing, video Processing, cryptography, molecular dynamics and computational fluid dynamics. We evaluate the FPGA based application acceleration with the emerging multicore architectures from the perspectives of price/performance and performance/watt. Specific contemporary reconfigurable computing systems are examined to identify existing system limitations and to highlight opportunities for research in dynamic and partial configuration areas.

Enrollment Requirements

Course Texts

Reconfigurable Computing: The Theory and Practice of FPGA-Based Computation, Scott Hauck, André DeHon, Morgan Kaufman, 2007.

Other reading material will be either presented in the class or available as online papers.

Schedule

Assessment

• Homework: 3-5 assignments
• Project: 1 project
• Exams: 1 midterm exam
• Typical grading policy: 20% midterm, 50% project, 20% homework, 10% participation

Probability, random variables, stochastic processes, correlation functions and spectra with applications to communications, control and computers.