Pinaki Mazumder co-authors new book: Memristive Computing

Pinaki Mazumder, professor of Electrical and Computer Engineering, co-authored Memristive Computing, a new book that explores research directions that use memristors for neuromorphic computing and artificial intelligence (AI). It is published by Springer and Science Press Beijing.

“The main benefit of this work is that we synthesized decades of memristor technology and brought it to cryptography, signal processing, adaptive filtering, and so on,” said Mazumder. “This book is very solid in mathematical foundation and there is a confluence of multiple research areas like materials science, computer science, device physics—so this book provides the impetus for doing more interdisciplinary fundamental research.”

Memristors are circuit elements that store data in a tunable charged state, allowing for efficient memory storage, reduced energy use, and enable circuits to resemble networks of neurons, the cells that process information in the brain. Mazumder has been working on research relating to memristors and brain-like (neuromorphic) computing since the late 1980s and early 90s.

At that time, the research that Mazumder was doing on “self-healing” very large scale integration (VSLI) circuits, cellular neural networks for image and video processing, reinforcement learning for applications like maze searches, and spike timing dependent plasticity—all breakthroughs in neuromorphic computing—was focused on connecting design vision to actual devices.

Yongbin Yu, associate professor at the University of Electronic Science and Technology of China and co-author of the book, collaborated with Mazumder as a visiting scholar at the University of Michigan in 2013-2014.

“Our approach was mostly targeted to engineering applications with realistic implementation,” said Mazumder, “and this work has been published in previous books: Genetic Algorithms in VLSI Design and Neuromorphic Circuits for Nanoscale Devices.”

“Yongbin went one step further, to consider things that were thought to be impractical.”

These atypical and revolutionary applications of memristor technologies include encryption and decryption of data, as well as the modeling and analysis of collective behavior in fish, birds, ants, and even artificial immune systems—problems that are classically difficult to solve.

Memristive Computing consists of eight chapters, including “Memristor and its Emulator,” “Chaotic Circuits Based on Memristor,” “Memristor-based En/Decryption System,” “Filter Design Based on Memristive Family,” “Memristive Filter for Signal Processing,” “Memristor-Based Swarm Intelligence,” “Dynamic Analysis of Memristive Neural Network,” and “Memristor-Based Neural Network and its Application.”

Many of these chapters were written by students in Yu’s research group. Yu also invited Mazumder to contribute his decades of expertise on memristive computing. The book is intended to encourage researchers across engineering, physics, and neuroscience to collaborate on the application of memristor devices to further challenging and non-Boolean problems at the cutting edge of computing.

“Engineering innovations are like restaurant dishes that take a long time to cook at low temperatures,” said Mazumder. “Starting in 1987 I incorporated genetic algorithms into VLSI design, then neuromorphic computing for self-healing VLSI design, then memristive computing for biological limits and information processing–and this work is a level of complexity beyond that.”

Memristive Computing is available for purchase via Springer and Amazon.