ECE Departmental Seminar
(Inching) Towards Computational Creativity: Using Insight from Cognitive Psychology for Data Driven Electronic Design Automation
Prof. Alex Doboli
Light Engineering 250
Summary: This talk will summarize our research efforts on devising next-generation electronic design automation algorithms expected to mimic human design creativity to some degree. Current automation methods are arguably geared to address cumbersome yet low intelligence activities. Still, the ongoing explosion in knowledge availability, on one side, and the need for more intelligent computing (e.g., superior autonomy, adaptation and customization), on the other side, creates intriguing opportunities to expand automated processing with new capabilities, such as human-specific creativity. We will overview our techniques for multi-view representation of design knowledge, design knowledge clustering, differential learning, supervised and unsupervised building block identification, causality extraction, design strategy mining, and data driven computational reasoning strategies. The methods are a new approach for tackling circuit and system topology design, often considered the “Holy Grail” in electronic design automation. While synthesis of analog circuits has been the main application domain for our work, the presented techniques can be extended for other areas too, like domain knowledge summarization, mining patterns in dynamic communities, healthcare, human-in-the loop applications, organizational design, strategic decision making, and human-machine interface design. This work was supported since 2009 by two NSF grants.
Bio: Alex Doboli received his engineering degree in Computer Science and Engineering (valedictorian), from “Politehnica” University Timisoara, Romania, and the “Doctor” degree in Computers from “Politehnica” University Timisoara in 1997. In 2000, he was awarded the Ph.D. degree in Computer Engineering from University of Cincinnati, Cincinnati, OH. He was a visiting researcher at Linkoping University, Sweden in 1994, 1995, and 1996. In Fall 2000 he joined the Department of ECE where he is now a Professor. Dr. Doboli's research is in Computer-Aided Design (CAD) of analog and mixed-domain systems and networks of systems, data driven algorithms, Cyber-Physical and embedded systems, and design methodologies for engineering innovation. He has published over 150 papers in peer-reviewed journals and conference proceedings, and has advised 11 graduated Ph.D. students. He co-authored with Dr. E. Currie the textbook “Introduction to Mixed-Signal, Embedded Design”, Springer, 2010. He received the 3rd prize at the National Math Student Contest, Romania (1986) and the “Traian Lalescu” Award for best student in mathematics from “Politehnica” University Timisoara (1987).