报告题目：Hardware SupportArchitectures and Implementation Methods for Effective Embedded AI
报告人：加拿大蒙特利尔大学工学院,Yvon Savari,, IEEE Fellow
摘要：After introducingselected elements of the former research contributions of Professor Savaria,this talk will focus on contributions of the DNN2GATE/CNN2GATE. This project isfinancially supported by the iVado AI research center that will be briefly outlined.The talk will introduce some recently introduced AI-ASICs, stressing some oftheir outstanding performances but also some of the challenges related to theirdevelopment and to their potential use in custom applications. This motivatesour DNN2GATE project that tries to get a great deal of the benefits of embeddedAI-ASICs using FPGA technology. It will specifically report on recent resultsrelated to developing a general framework to implement Deep ConvolutionalNeural Networks. The talk finally makes the case for using FPGAs as a vectorfor demonstrating powerful embedded AI hardware accelerators and for using themin white box deployable and certifiable embedded AI targets for criticalapplications
报告人简介：YvonSavaria FIEEE received the B.Ing. and M.Sc.A in electrical engineering fromPolytechnique Montreal Canada in 1980 and 1982 respectively. He also receivedthe Ph.D. in electrical engineering in 1985 from McGill University. Since 1985,he has been with Polytechnique Montreal, where he is currently professor anddirector of the Microelectronics Research Group in the department of electricalengineering.
Hehas carried work in several areas related to microelectronic circuits andmicrosystems such as testing, verification, validation, physical designmethods, defect and fault tolerance, effects of radiation on electronics, CADmethods, reconfigurable computing and applications of microelectronics totelecommunications, aerospace, image processing, video processing, radar signalprocessing, and digital signal processing acceleration. He is currentlyinvolved in several projects that notably relate to wireless sensor networks,virtual networks, machine learning, computational efficiency and applicationspecific architecture design. He used artificial neural networks and a widerange of hardware acceleration techniques to implement effective hardwaresupported signal processors over the past 30 years. He holds 16 patents, haspublished more than 150 journal papers and more than 480 conference papers.
Hehas been working as a consultant or was sponsored for carrying research by morethan 20 companies or research organizations. He is a member of the RegroupementStratégique en Microélectronique du Québec (Canada) (RESMIQ), of the Ordre desIngénieurs du Québec (OIQ - Canada), and is a member of CMC Microsystems Board.He wasco-founder of two high-tech spin-offs and an early collaborator to severalothers. He also received in 2006 a Synergy Award of the Natural Sciences andEngineering Research Council of Canada for his work with LTRIM.