Analog Hardware, for Solving the Hardest Problems in Computer Science
About 70 years ago, analog computing was regarded as having equal prospects as digital computing. Operational amplifiers could provide analog differentiation and integration functions. Nonetheless analog computing disappeared, being unable to provide the precision and dynamic range required for solving real problems.
The emergence of Deep Learning has been accompanied by the realization that only modest precision is sufficient for the inference step. This has taken us from regular Floating Point, to half-precision (16 bits), to quarter-precision, and with some difficulty even single-bit precision. The race is on for specialized hardware accelerators, whose acronyms have transitioned from CPU->GPU->TPU->IPU. For example, eight-bit precision analog can provide analog matrix multiplication in Deep Learning accelerators, which is now being pursued commercially.
The speaker will examine three different potential forms of analog computing:
(a) analog matrix multipliers for Deep Learning,
(b) analog, not digital simulated annealing for solving Ising type problems, and
(c) adiabatic computing (classical not quantum), also for solving Ising type optimizations.
One of the limitations for Ising problems is that the analog couplings demand precision that grows with problem size. It appears that, already at one percent analog precision, interesting Ising problems can be addressed.
About the speaker
Prof. Eli Yablonovitch received his PhD in Applied Physics from Harvard University in 1972. He has worked at Bell Telephone Laboratories, Exxon and Bell Communications Research, where he was Director of Solid-State Physics Research and began his work in photonic crystals. Before joining the University of California, Berkeley in 2007, where he is currently James & Katherine Lau Chair in Engineering, Prof. Yablonovitch had also taught at Harvard University and the University of California, Los Angeles. He was awarded a Doctor of Engineering honoris causa from the Hong Kong University of Science and Technology in 2011.
Prof. Yablonovitch's work has covered a broad variety of topics: nonlinear optics, laser-plasma interaction, infrared laser chemistry, photovoltaic energy conversion, strained-quantum-well lasers, and chemical modification of semiconductor surfaces. Currently his main interests are in optoelectronics, high speed optical communications, high efficiency light-emitting diodes and nano-cavity lasers, photonic crystals at optical and microwave frequencies, quantum computing and quantum communication.
Prof. Yablonovitch was elected Member of the US National Academy of Engineering, the US National Academy of Sciences and the Royal Society of London (foreign member). He is also a Fellow of the Institute of Electrical and Electronics Engineers (IEEE), the Optical Society of America, the American Physical Society, the American Academy of Arts and Sciences, and the US National Academy of Inventors. He has been awarded the Benjamin Franklin Medal in Electrical Engineering (2019), the IEEE Edison Medal (2018) and Photonics Award (2012), the William R. Cherry Award (2017), the Oliver E. Buckley Condensed Matter Physics Prize (2016), the Isaac Newton Medal (2015), the Harvey Prize (2012), the Julius Springer Prize for Applied Physics (2001), the Institution of Engineering and Technology Mountbatten Medal (2010), the R.W. Wood Prize (1996), the William Streifer Scientific Achievement Award (1993) and the Adolph Lomb Medal (1978).