Prof. Bahram Jalali: Accelerating computing with trainable nonlinear Schrödinger equation

This is a hybrid seminar!

https://illinois.zoom.us/j/6585262376?pwd=kEldRS8hjgHqJ92lZNVbBqBduI9r6u.1

Meeting ID: 658 526 2376
Password: 861201

Abstract:

We introduce a new framework where master equations of physics are converted into general purpose compute “networks” that learn complex patterns in data. The first implementation utilizes the Nonlinear Schrödinger Equation (NLSE) and has two implementations: the Nonlinear Schrödinger Kernel (NSK), and the Nonlinear Schrödinger Network (NSN). The first is a physical hardware compute accelerator using femtosecond pulses for data acquisition and computing, leveraging nonlinear optics as an analog kernel to enhance data classification accuracy. While optimizable via phase encoding in hardware, it faces limitations due to the rigid properties of the physical optical medium which does not lend itself to easy adaptation and training. To overcome these limitations, our second implementation, the Nonlinear Schrödinger Network (NSN), operates in the numerical domain. NSN is a general-purpose trainable model for learning complex memory and nonlinear behavior in data. It offers a more interpretable and parameter-efficient alternative to traditional black-box AI models such as convolutional neural networks (CNNs). It achieves comparable or superior accuracy in time series classification tasks with significantly fewer parameters. This property holds promise for solving the “memory wall” problem facing AI clusters.

Biography:
Bahram Jalali is a Distinguished Professor Emeritus at the ECE Department at UCLA. He was the Founder and CEO of Cognet Microsystem, a pioneering CMOS fiber-optic integrated circuits company acquired by Intel (NASDAQ: INTC) in 2001. A second company spun off from his research to commercialize his radio-frequency based fluorescence camera invention for cell biology was acquired by BD Biosciences (NYSE:BDX) in 2017. He was inducted into the National Academy of Engineers (NAE) and the National Academy of Inventors (NAI) for his pioneering contributions to silicon photonics and optical instruments. He is a fellow of APS, IEEE, Optica, SPIE, and AIMBE, and one of the top 10 authors in the journal Nature Photonics.  His work has been recognized in the Scientific American Top 50 and MIT Technology Review Top10 and has received the Optical Society of America’s (Optica) R.W. Wood Prize and the IEEE’s Aaron Kressel Award among others. Bahram began his career at the age of 25 as a Member of Technical Staff in the Physics Research Division of Bell Laboratories in Murray Hill, N.J., after completing his Ph.D. in Applied Physics at Columbia University. He has more than 50 US patents.