Скачать или смотреть How Can Neuroscience Aid in Designing a Mortal Computer | Ayon Borthakur | ATLS-24 | AAII

The demand for intelligent computing is on the rise. But for numerous practical applications that involve real-time interactions, current AI algorithms, and the associated computing architectures still cannot match the performance of the brain. For instance, unlike biological neural networks, artificial neural networks (ANNs) tend to forget previously learned information upon learning new information (catastrophic forgetting). Moreover, today’s computers used for the implementation of ANNs are slow and energy-consuming due to factors such as the physical separation of computing units and memory, unlike biological neural networks where computation and memory are co-localized. Accordingly in this research, we take neuroscience inspiration and work towards designing a mortal computer.
In this talk, Ayon will describe the properties of such a mortal computer and showcase selected neuroscience-inspired AI design results. He will present EPLff - inspired by the external plexiform layer of mammalian olfaction, which learns from a few shots without catastrophic forgetting. Next, he will discuss Sapinet - a multilayer recurrent neural network for learning in the wild optimized for implementation in Intel Loihi.

"I am an Assistant Professor in the Department of Artificial Intelligence at IIT Hyderabad. Previously, I worked at Innatera Nanosystems in the Netherlands as a Senior Neuromorphic Engineer-Machine Learning. At Innatera, I worked on marrying deep learning and analog computing towards ultra-low power and latency radar target recognition, which resulted in multiple patent applications. Before that, I completed my Ph.D. at Cornell University, USA. As part of my Ph.D., I worked on neuroscience-inspired Artificial Intelligence for learning in the wild, focusing on its implementation in neuromorphic chips such as Intel Loihi. My Ph.D. work is part of an international patent sponsored by Cornell. I completed my Bachelor of Technology in Electrical Engineering from IIT Dhanbad."

https://borthakurayon.github.io/