Selected Publications

This paper deals with a progressive recovery problem under limited resources in networks with VNFs, where some dependent network layers exist. We prove the NP-hardness of the progressive recovery problem and approach the optimum solution by introducing DeepPR, a progressive recovery technique based on deep reinforcement learning. Our simulation results indicate that DeepPR can obtain 98.4% of the theoretical optimum in certain networks. The results suggest the applicability of Deep RL to more general progressive recovery problems and similar intractable resource allocation problems.
DeepPR, 2019

Recent & Upcoming Talks

Convex Functions for Reinforcement Learning
Apr 23, 2019 1:15 PM
Robust Optimization with Applications in Networking
Apr 23, 2019 11:30 AM
Experience Driven Networking: A Reinforcement Learning Based Approach
Mar 12, 2019 11:30 AM

Recent Posts

Interpretable neural networks are currently nothing more then a pipe dream. However, if we were to make these machines even partially interpretable, maybe we could learn more about how they work. [ Placeholder ] Visually interpretable is a broad term. Here is a T-SNE embedding on the MNIST dataset, where 0’s, 1’s, etc. are clustered in their specified classes: The Network We wish to create a neural network that not only gives us a class prediction given an input image, but also a 3-dimensional embedding within a defined space.

CONTINUE READING

Projects

RuTroll

Chrome extension which highlights tweets likely to come from a Russian troll farm. Classifier trained on fivethirtyeight open access dataset. HackHarvard 2018.

Intel Innovate FPGA

Semi-finalist of Innovate FPGA. Accelerating a neural network with an FPGA (DE10-Nano) for detection of bicyclists at intersections.

Green Raccoon

A react-native application which takes a picture and tells you if something is recyclable or not. react-native, google-cv-api, snack.expo

Deep Margins

Attempting to increase decision margins in neural networks through techniques similar to margin maximization in SVMs. Mentored by Prof. Nicholas Ruozzi, UTD. Python, Tensorflow, open-cv, alot of bash.

PhysVR

Developed a unobtrusive method to virtually display a physical object for collision avoidance in virtual reality. Work done in the Future Immersive Virtual Environments Lab (Mentored by Dr. Ryan P. McMahan) through the Clark Summer Research Program. Unity3D, C#, and the HTC VIVE headset.

MyUTD

Cross-platform mobile application that tracks and displays the locations of on-campus transportation systems realtime. QT, QML, JavaScript, and C++.