Bharat Prakash

I am a PhD student in CSEE at University of Maryland, Baltimore County (UMBC), where I am currently working on Machine Learning and Deep Reinforcement Learning.

I did my MS also at UMBC and BE at University of Pune. And then, I spent time at eBay/PayPal as a software enginner where I worked on PayPal Credit and PayPal Working Capital.

Email  /  CV  /  LinkedIn  /  GitHub


My research interests include deep learning, computer vision and deep reinforcement learning (RL). Currently I am working on some research projects in safe RL and human in the loop RL.

Improving Safety in Reinforcement Learning using Model-Based Architectures and Human Intervention
Bharat Prakash, Mohit Khatwani, Nicholas Waytowich, Tinoosh Mohsenin
Accepted, FLAIRS, 2019

A hybrid method for reducing the human intervention time by combining model-based and model-free approaches and training a supervised blocker to improve sample efficiency while also ensuring safety. We evaluate these methods on various grid-world environments using both standard and visual representations and show that our approach achieves better performance in terms of sample efficiency, number of catastrophic states reached as well as overall task performance compared to traditional model-free approaches.

Representation learning by solving auxiliary tasks on Xray images
Bharat Prakash

Learning image representations on unannotated Chest Xray images using the method described in Noroozi and Favaro to gain improvements in classification tasks. Here we they use the pretext task of solving jigsaw puzzles to pre-train the convolutional neural network. Chest x-ray images from the x-ray14 database were used.

Improving Grammatical Error Correction Using Multi-Task Learning
Bharat Prakash, Ashwinkumar Ganesan, Sarthak Mehta, John Cellozi, Frank Ferraro
Poster Presentation at MASC-SLL-2018

We propose a multitask learning (MTL) approach to design a grammatical error correction system that uses minimal parallel text for training by adding auxiliary tasks that aid the main GEC training objective. The proposed experiments include adding part of speech tagging and language modeling as these additional auxiliary tasks.

Other Projects

Colorizing monochrome photographs
Bharat Prakash, Harsh Shrivastava, Deepanjan

Using deep neural networks to colorize monochrome photos. Course project, Data Science.

University FAQ ChatBot
Bharat Prakash, Abhay Kashyap

Facebook chatbot to answer basic questions about the university. Natural language understanding, NER, information retrieval.

IVR Payment
Bharat Prakash, Abhay Kashyap

Developed an IVR system to process payments using PayPal APIs, Asterisk on Raspberry Pi. Internal hackathon at PayPal.

(wesbite code from this guy)