Applied CBR to knowledge graph completition with competitive results. Demonstrated usage instances where it outperforms existing embedding-based methods. Working to extend functionality to raw text.
Data Center Energy Portfolio Optimization
Proposed, implemented and tested reinforcement learning and imitation learning based agents for data center battery controllers. Outlined future work directions.
Progressively Balanced Multi-class Neural Trees
Proposed and tested an entropy impurity based objective function for incorporating a learnable perceptron into the decision tree framework. Demonstrated that the learned classifier achieves comparable accuracy with fewer test time computations than an MLP
Commercial Segmentation in Television Stream
Attempted segmentation of commercials in television stream through audio-visual feature engineering and application of sequence classifier.
Quora Question Pairs
Trained a Siamese Gated Recurrent Unit (GRU) RNN over sentence pairs to detect duplicate questions, securing a position in the top 25% among 3000+ teams on Kaggle.
Bufferless Routeing in Network-On-Chip
Designed a network traffic simulator to study on-chip interconnection networks composed of bufferless routers.
Autonomous Intelligent Robot
Implemented simultaneous localization and mapping (SLAM) and visual odometry to create a robot capable of mapping and navigating its environment.