Big Data Analytics in VLSI Design Automation and Test-Principles, Challenges, and Promises
Speaker: Li-C. Wang (Professor, University of California, Santa Barbara)
Topic: Big Data Analytics in VLSI Design Automation and Test-Principles, Challenges, and Promises
Date : 09:00～16:30 Friday 24 April. 2015
Place : 102 Delta Building
Summary:
Electronic Design Automation (EDA) and Test have become major application areas for data analytics in recent years. In design and test processes, tremendous amounts of simulation and measurement data are produced and collected. These data present opportunities for applying data analytics.
“Big Data” in this tutorial is viewed as an application utilizing data that can be continuously collected over time. In contrast, application that is restricted to data from a limited period of time is considered as a “Small Data” application. With this differentiation in mind, in the tutorial I will teach the principles of various data mining approaches and explain “Big/Small Data” application examples to illustrate how these approaches can be applied in EDA and test in practice. I will then explain the working principles of specific learning techniques in supervised learning, unsupervised learning and rule learning. I will present examples as how to formulate an EDA/test problem to facilitate the application of these learning techniques, in areas such as functional verification, simulation trace analysis, layout analysis, timing analysis, design-silicon correlation, Fmax prediction, delay testing, test cost reduction, diagnosis, and customer return analysis. The applications will be divided into two essential paradigms. In the knowledge discovery paradigm, the objective of data analytics is to discover interpretable and actionable knowledge. I will highlight the challenges in this paradigm and provides a summary of the applications. In the prediction paradigm, the objective is to predict future trends or properties based on data available so far. The challenges in the second paradigm will also be highlighted with a summary of applications. Experience of developing a practical data mining flow will be presented and promises of applying data analytics in practice will be demonstrated, through experimental results in selective applications based on industrial settings.
The tutorial is intended for students, engineers, and researchers who are interested in understanding data analytics and how data mining techniques can be applied in design automation and test in practice. Tutorial covers the knowledge for someone interested in pursuing a learning approach in their respective applications and/or interested in assessing the potential of data mining that can be brought to their research and development efforts.
Keywords: Computer-Aided Design, Data Mining, Machine Learning, Test, Verification, Validation
About the presenter:
Li-C. Wang is professor of ECE department at UC-Santa Barbara. He received Ph.D. in 1996 from University of Texas at Austin. He was a senior CAD software technical staff member at PowerPC Design Center, Motorola from 1996 to 2000, where he led various projects for PowerPC microprocessor test and verification. Dr. Wang received best paper awards from DATE-1998, IEEE VTS-1999, DATE-2003, VLSI DAT 2008 and 2011. He received the Technical Excellence Award from Semiconductor Research Cooperation (SRC) in 2010 for contribution on developing data mining technologies in the areas of test and validation. He co-founded the IEEE Microprocessor Test and Verification (MTV) Workshop, and is currently the program co-chair. He is currently serving or had served as technical PC member for various workshops and conferences including ITC, VTS, ICCAD, DATE, DAC, ISQED, HLDVT, ITSW, DATA, ATS, ICCD, VLSI-DAT, etc and is currently serving as the general co-chair for VLSI-DAT. He is an associate editor of IEEE Trans. on CAD and also guest editors of a number of IEEE D&T, JETTA, and ACM TODAES special issues. From 2005, his research group has published more than 70 papers on the topics related to data mining and machine learning in test, verification and validation. In the last two years, he had given six tutorials on data mining and had been invited to lecture on data mining in DAC, ICCAD, ASP-DAC and ISPD.
Topic: Big Data Analytics in VLSI Design Automation and Test-Principles, Challenges, and Promises
Date : 09:00～16:30 Friday 24 April. 2015
Place : 102 Delta Building
Summary:
Electronic Design Automation (EDA) and Test have become major application areas for data analytics in recent years. In design and test processes, tremendous amounts of simulation and measurement data are produced and collected. These data present opportunities for applying data analytics.
“Big Data” in this tutorial is viewed as an application utilizing data that can be continuously collected over time. In contrast, application that is restricted to data from a limited period of time is considered as a “Small Data” application. With this differentiation in mind, in the tutorial I will teach the principles of various data mining approaches and explain “Big/Small Data” application examples to illustrate how these approaches can be applied in EDA and test in practice. I will then explain the working principles of specific learning techniques in supervised learning, unsupervised learning and rule learning. I will present examples as how to formulate an EDA/test problem to facilitate the application of these learning techniques, in areas such as functional verification, simulation trace analysis, layout analysis, timing analysis, design-silicon correlation, Fmax prediction, delay testing, test cost reduction, diagnosis, and customer return analysis. The applications will be divided into two essential paradigms. In the knowledge discovery paradigm, the objective of data analytics is to discover interpretable and actionable knowledge. I will highlight the challenges in this paradigm and provides a summary of the applications. In the prediction paradigm, the objective is to predict future trends or properties based on data available so far. The challenges in the second paradigm will also be highlighted with a summary of applications. Experience of developing a practical data mining flow will be presented and promises of applying data analytics in practice will be demonstrated, through experimental results in selective applications based on industrial settings.
The tutorial is intended for students, engineers, and researchers who are interested in understanding data analytics and how data mining techniques can be applied in design automation and test in practice. Tutorial covers the knowledge for someone interested in pursuing a learning approach in their respective applications and/or interested in assessing the potential of data mining that can be brought to their research and development efforts.
Keywords: Computer-Aided Design, Data Mining, Machine Learning, Test, Verification, Validation
About the presenter:
Li-C. Wang is professor of ECE department at UC-Santa Barbara. He received Ph.D. in 1996 from University of Texas at Austin. He was a senior CAD software technical staff member at PowerPC Design Center, Motorola from 1996 to 2000, where he led various projects for PowerPC microprocessor test and verification. Dr. Wang received best paper awards from DATE-1998, IEEE VTS-1999, DATE-2003, VLSI DAT 2008 and 2011. He received the Technical Excellence Award from Semiconductor Research Cooperation (SRC) in 2010 for contribution on developing data mining technologies in the areas of test and validation. He co-founded the IEEE Microprocessor Test and Verification (MTV) Workshop, and is currently the program co-chair. He is currently serving or had served as technical PC member for various workshops and conferences including ITC, VTS, ICCAD, DATE, DAC, ISQED, HLDVT, ITSW, DATA, ATS, ICCD, VLSI-DAT, etc and is currently serving as the general co-chair for VLSI-DAT. He is an associate editor of IEEE Trans. on CAD and also guest editors of a number of IEEE D&T, JETTA, and ACM TODAES special issues. From 2005, his research group has published more than 70 papers on the topics related to data mining and machine learning in test, verification and validation. In the last two years, he had given six tutorials on data mining and had been invited to lecture on data mining in DAC, ICCAD, ASP-DAC and ISPD.