Last updated on: 16th June, 2023

Hi! My name is Utkarsh Jain, and I’m currently pursuing a graduate degree in Computer Science at the University of California San Diego. My research focuses on the development of biologically inspired Convolutional Neural Networks (CNNs) under the guidance of Prof. Gary Cottrell and Prof. Virginia De Sa. This exciting field lies at the intersection of Cognitive Science and Computer Science, and it has yielded numerous fascinating discoveries. One particularly intriguing finding is that as CNNs improve their performance in image categorization tasks, their internal representations become increasingly similar to those found in the human brain. Currently, I am investigating whether further aligning CNN representations with brain-like patterns can enhance their performance.

Prior to joining UC San Diego, I worked as a researcher at the Open Source Intelligence Lab in the Indian Institute of Technology Guwahati under the guidance of Prof. Sanasam Ranbir Singh. During my time there, I delved into the application of Natural Language Processing (NLP) and Graph Neural Networks to predict the home locations of Twitter users. This research had significant implications for various online services, such as targeted advertising, opinion mining, and event detection. It was my first foray into NLP, and I thoroughly enjoyed the experience of learning new concepts and collaborating with talented individuals.

I completed my Bachelor of Technology in Computer Science and Engineering at the Indian Institute of Technology Mandi. As part of a semester exchange program, I had the privilege of studying at Aalto University in Finland, where I had the opportunity to explore several captivating courses. It was during this period that I discovered my passion for mathematics and made the decision to pursue a career in Machine Learning. Upon returning from Aalto University, I worked with Prof. Aditya Nigam on investigating the applications of Reinforcement Learning (RL) in the task of Image Classification. Our research revealed that while RL performed equally well as CNNs in classifying images, it is significantly less time-efficient. Also, it is tricky to convert a supervised problem into a reward-based problem. How do you design an efficient reward function for image classification? And what does the environment look like?. Furthermore, I had the privilege of collaborating with Prof. Mousa Marzband at Northumbria University, United Kingdom, where we applied time-series modeling to weather now-casting.

Over the past few years, I have experienced profound moments of enlightenment that have shaped my life’s trajectory. I discovered my passion (or did I develop it?) which has ignited a deep sense of purpose within me, and I am now determined to pursue it wholeheartedly. I have come to recognize the inherent unpredictability that permeates our existence and the significant influence of circumstance and luck in shaping our paths. My ongoing journey has been a tremendous source of learning, with one of the most valuable lessons being the art of embracing my humanity.