Research
Deep Learning | Natural Language | Computer Vision | Multi-agent Systems
Embedded Files
WildMe - 2022
Data-Efficient Image Segmentation
Data-Efficient Image Segmentation
Developed deep learning based method of automatically segmenting animal images for animal identification pipeline. Significantly reduced cost of annotation and increase segmentation performance with self-supervision and weak-supervision training scheme.
Northwestern University - 2022
Multi-Agent Reinforcement Learning
Multi-Agent Reinforcement Learning
Researched plan-sharing and collaboration among deep RL agents. Developed RL ensemble learning method where agents learned to leverage sophisticated plan sharing strategies over simple traditional ensembeling methods.
Northwestern Image Video Processing Lab (IVPL) - 2021
Video Super Resolution w/ Vision Transformers
Video Super Resolution w/ Vision Transformers
Developed new deep learning architectures for video super resolution (VSR) based on vision transformers (ViT) and CNNs. Experimentation done on Vimeo-90K dataset.
Northwestern University - 2021
Narrative Generation Architecture
Narrative Generation Architecture
Collaborated with Northwestern C3 Lab to develop narrative generation architecture based on LLM and narrative planning system. The objective was to combine traditional AI planning engines, narrative frames, and LLMs (GPT-2) to generate a coherent narrative script based on input character and plot parameters. This was at a time when the best LLMs failed at generating long coherent structured narratives.
Northwestern University - 2021
Novelty Generation w/ Neural Cellular Automaton
Novelty Generation w/ Neural Cellular Automaton
Researched methods for measuring and generating novelty and open-endedness using cellular automaton. Trained with genetic algorithms and as neural cellular automaton (NCA).
Benedictine University - 2020
Adversarial Defense
Adversarial Defense
Researched several variational auto-encoder (VAE) based methods of defending image models against adversarial attacks by reconstructing input images to remove adversarial perturbations.
Other Research Projects
Shorter independent or university research topics
Sentence-embedding Narrative Structure Analysis (2022)
Neural Radiance Fields (2022)
Deep Reinforcement Learning Portfolio Trading Agent (2021)
Chicago Police Interaction Network Analysis (2020)
AE + GAN Style Transfer (2018)
Page updated
Google Sites
Report abuse