Chui Yi (Iris) Liu

I am a Master's student in Computer Science (AI focus) from University of Southern California, where I am a research assistant at Robotics Embedded Systems Laboratory, advised by Gaurav Sukhatme.

Lately, I've worked on garage at RESL and investigated in topics related to reinforcement learning and robotics. I also serve as the Head of Programs at USC Graduates Rising in Information and Data Science (GRIDS).

I am interested in the indisciplinary areas of artificial intelligence and machine learning. In partcular, I'm interested in computer vision, deep learning, and reinforcement learning.

CV / Email / GitHub / LinkedIn / Blog

Projects

These include coursework, side projects and unpublished research work.

	A Continuous Indicator of Food Environment Nutritional Quality Iris C. Liu, Kayla de la Haye, Andrés Abeliuk, Abigail L. Horn KDD, 2021 paper / website We introduces a summarizing indicator of restaurant nutritional quality that exists along a continuum, and which can be applied at scale to make distinctions between diverse restaurants within and across categories of food outlets.
	Maverick: Multi-drone collaboration USC CS527 : Applied Machine Learning for Games 2020-11-11 website We train a PPO that allows multiple drone agents to perform collaborative tasks including pursuit & evasion, dynamic obstacle scenarios in a 3D physics based simulator.
	Race On - self-driving car competition competition 2020-03-12 code With Guanyang Luo, Shubham Nagarkar and Posha Dave, we assembled a car from scratch and programmed the control algorithm of a self-driving car to complete a lap in 9.6 seconds.
	MICCAI - Retinal Fundus Glaucoma Challenge workshop 2018-06-01 By inputing the pre-processed fundus images to DenseNet-161 for segmentation to calculate the cup-to-disc ratio for diabetic retinopathy and glaucoma screening.
	Short Text Classification with Deep Neural Networks: An Experimental Analysis Iris C. Liu CityU Best Student Paper, 2017 paper / code / slides In the context of short texts, LSTMs is applied to perform text classification. This method demonstrates that using temporal LSTMs can achieve remarkable performance in capturing syntactic and semantic meaning of short-texts without expensive feature engineering and understanding text structures.
	Man2Hero: A game AI with javascript CityU CS4386 : Game AI Design 2016-06-01 video / code ‘Man 2 Hero’ is a single-player action-puzzle PC game with game AI implemented at each level. The highlight of the game is on the AI engine design which includes rule-based fuzzy logic and finite state machine logic (FSM).