Fast-Converging Deterministic Approaches for Stochastic Computing

شروع رویداد
چهارشنبه ۱۲ دی ۹۷ ۱۲:۳۰
پایان رویداد
چهارشنبه ۱۲ دی ۹۷ ۱۴:۰۰
مکان رویدادتهران
اضافه به تقویم
Fast-Converging Deterministic Approaches for Stochastic Computing
مهلت ثبت‌نام برای این رویداد به پایان رسیده است.
جزئیات
مهلت ثبت‌نام
قیمت (تومان)
تعداد

توضیحات

 

موسسه بین‌المللی توسعه دانش فردای ایرانیان با همکاری دانشگاه تهران برگزار می‌کند:

 

سمینار تخصصی 

Fast-Converging Deterministic Approaches for Stochastic Computing

 

چکیده سمینار

Stochastic computing (SC), a paradigm first introduced in the 1960s, has received considerable attention in recent years as a potential paradigm for emerging technologies and “post-CMOS” computing. Logical computation is performed on random bitstreams where the signal value is encoded by the probability of obtaining a one versus a zero. Implementing complex functions with simple hardware (e.g., multiplication using a single AND gate), tolerating soft errors (i.e., bit flips), and progressive precision are the primary advantages. The paradigm, however, suffers from high latency and poor precision. The results are always somewhat inaccurate due to random fluctuations. High latency also translates to high energy consumption. We show that randomness is not a requirement for this computational paradigm. We present three deterministic approaches for the computation: relatively prime stream lengths, rotation, and clock division. These reduce the area, greatly reduce the latency (by an exponential factor), and produce completely accurate results. We further introduce two fast-converging, scalable deterministic approaches to SC based on low-discrepancy sequences. The results are completely accurate when running the operations for the required number of cycles. However, the computation can be truncated early if some inaccuracy is acceptable. Experimental results show that the proposed approaches significantly improve both the processing time and area-delay product compared to prior approaches.

 

با حضور

M. Hassan Najafi

Assistant Professor

University of Louisiana

چهارشنبه 12 دی ماه 1397، از ساعت 12:30 تا 14:00

تهران، خیابان کارگرشمالی، پردیس دانشکده‌های فنی، دانشکده مهندسی برق و کامپیوتر دانشگاه تهران، ساختمان شماره 2، اتاق 803

 

سخنرانان

دکتر محمدحسن نجفی

دکتر محمدحسن نجفی

استادیار دانشگاه لوئیزیانا

Dr. M. Hassan Najafi received the B.Sc. degree in Computer Engineering-Hardware from the University of Isfahan, Isfahan, Iran, the M.Sc. degree in Computer Architecture from the University of Tehran, Tehran, Iran, and the Ph.D. degree in Electrical Engineering from the University of Minnesota, Twin Cities, USA, in 2011, 2014, and 2018, respectively. He is currently an Assistant Professor with the School of Computing and Informatics, University of Louisiana at Lafayette, Louisiana, USA. His research interests include stochastic and approximate computing, unary processing, computer architecture, low power VLSI design, and designing fault-tolerant systems.
Dr. Najafi’s research establishes some counterintuitive, yet fundamental, new design methodologies for designing digital stochastic systems. Conceptually, his works challenge the limitations of bit-stream-based computing; practically, they provide a means for designing significantly smaller, faster, and energy-efficient embedded systems. In recognition of his research, he received the Doctoral Dissertation Fellowship from the University of Minnesota and the Best Paper Award at the 2017 35th IEEE International Conference on Computer Design (ICCD). His work on Polysynchronous Clocking was selected as the Feature Paper of the Month in the Oct 2017 Issue of the IEEE Transactions on Computers.

برگزارکنندگان

موسسه بین المللی توسعه دانش فردای ایرانیان

موسسه بین المللی توسعه دانش فردای ایرانیان

دانشگاه تهران

دانشگاه تهران

آدرس:تهران تهران، خیابان کارگرشمالی، پردیس دانشکده‌های فنی، دانشکده مهندسی برق و کامپیوتر دانشگاه تهران، ساختمان شماره 2، اتاق 803