The recent resurgence in neuromorphic computation comes in the wake of the traditional Von-Neumann computers reaching their performance limits. Unfortunately most of the efforts until now to develop neuromorphic hardware have been concentrated on using Si based Complementary Metal Oxide Semiconductor (CMOS) technology as the basic building block. To make neuromorphic circuits reach their true potential in terms of efficiency new technologies are needed for the neuromorphic circuit components i.e. the artificial synapses and neurons. In the first phase of my talk I will introduce some of the novel device technologies, especially resistive switching devices or “memristors” that can be used for neuromorphic applications. Resistive switches are one of the most popular candidates for artificial synapses due to their scalability, simple device structure, ease of 3D integration and high on/off ratios. However, significant challenges especially high power consumption and reliability issues remain that are needed to addressed. I will discuss about the strategies to develop forming-free low power resistive switches and reduction of device variabilities. Potential applications of resistive switches as well as other novel device technologies such as negative capacitance transistors as artificial neurons will also be discussed. The second part of the talk will involve discussion on neuromorphic networks with memristive arrays. We will discuss in particular about the concept of CMOL (CMOS + Molecular) circuits, advantages and practical implementation of hybrid 3D CMOS/memristor crossbar networks. Development of deep neural networks or “Perceptrons” with passive memristive crossbar arrays and future possibilities of all memristive neural networks will also be discussed. The third and final part of the talk deals with the possibility of creating bio-hybrid interfaces with biological neurons interfaced with artificial neural networks. Such hybrid interfaces can be hugely beneficial for developing basic understanding of neural connectivity as well as biomedical applications. Pathways to create such hybrid interfaces, practical design challenges and future applications will be discussed.
Bhaswar Chakrabarti is a post-doctoral scholar at the Institute for Molecular Engineering in the University of Chicago. He is broadly interested in the areas of electronic devices, oxide based high-k dielectric materials for non-volatile memories and neuromorphic computation. His current research is aimed at the investigation of novel material systems for the development of low power artificial synapses. Bhaswar has previously worked as a post-doctoral scholar (2014-2016) at the University of California,Santa Barbara where he was involved in the development of neuromorphic circuits with memristive devices. He was responsible for the first demonstration of a monolithic 3D CMOS-memristor hybrid integrated circuit with the capability to work as a high performance multiply-add engine. His other works at UCSB include development of a multi-layer deep neural network (perceptron) with memristive crossbar arrays. Bhaswar received his B. Tech in Radiophysics and Electronics from the University of Calcutta, Kolkata, India, in 2005 and M. Tech in Nanoscience and Technology from Jadavpur University, Kolkata, India, in 2007. He obtained his Ph.D. in Materials Science & Engineering from the University of Texas at Dallas, Richardson in 2013. His doctoral research primarily focused on the development of low power forming free resistive switching memory devices using high-k dielectrics and novel 2 dimensional electrode materials