Kong Ko Lun
Tagline:Year 4 Undergraduate Student at the City University of Hong Kong
Hong Kong
About Me
I am an undergraduate student passionate about research in AI and its applications in products, robotics, and other decision-making tasks. Over the past years, I have pursued my interest through developing AI applications during internships at a company in Science Park and Chow Sang Sang, as well as in research projects, forming and developing applications inspired by the latest research. My vision is to integrate research from different fields into AI research to build efficient, functional, and accurate AI agents that can be applied to autonomous robots in simulation and reality.
Presentations/Seminars
Navigation of Tendon-driven Flexible Robotic Endoscope through Deep Reinforcement Learning
Date: Jul 2024
Event name: IET HK YPEC (Young Professional’s Exhibition and Competition) 2024 .Location: City University of Hong Kong, Hong Kong .
Description:I presented the implementation details (endoscope simulation environment, reinforcement Learning training parameters (state, action, reward space)) and experiment results (episodic rewards, success rate given margin of error) at the YPEC 2024.
Using AI to Model Business Growth and Its Implications for Business
Date: Feb 2024
Location: Centre for Complexity and Complex Networks (CCCN), City University of Hong Kong .
Description:Statistical modeling of univariate data, like that of business growth data, could be complex and challenging. This talk aims to show the simplification of such a process by utilizing network sciences to retrieve a simple ordinary differential equation (ODE) to model growth and using artificial intelligence (AI) to extract helpful information from growth data by estimating the parameters of the differential equation. How business owners can use such info to grow their businesses will be discussed.
Patents
AN INTERACTIVE ELDERLY ASSISTANCE SYSTEM
Date: Jul 2023
Patent Number: HK30085199 .Status:Issued.
Inventors:Hong YANLEE Man Wah GloriaCHAN Yiu KwongZHANG YupeiGUO RunzhouKONG Ko LunDescription:Main contribution is the usage of Natural Language Processing (NLP) techniques for the chatting function of the system.
Projects
Deep Reinforcement Learning for Navigation of Soft Robotics
date: 2024Organization:Chinese University of Hong Kong (CUHK)
Description:This project was done in Prof. Hongliang Ren’s lab at the Chinese University of Hong Kong. Work completed in this project includes the creation of a simulation environment with a Tendon-Driven Robotic Endoscope (TDRE) suitable for training Reinforcement Learning (RL) agents and using Proximal Policy Optimization (PPO) for navigation with and without contact (free space). Details and results are presented at the Young Professional’s Exhibition and Competition (YPEC 2024).
Reasoning of Large Language Models for the Abstraction and Reasoning Corpus (ARC-AGI)
date: 2024Description:For this competition, my aim is to explore the reasoning capabilities of LLMs for challenging task-specific scenarios. The ARC-AGI corpus was used as the dataset, a small LLaMA model was finetuned using techniques such as low-rank, reasoning and internal self-correction. The model achieves ~75% on the training set and 2% on the hidden test set from the ARC 2024 competition, indicating that LLMs are able to solve difficult unseen problems despite overfitting.
Application of Scientific AI on Time Series Data
date: 2023Organization:City University of Hong Kong (CityU)
Description:This project is supervised by Prof. C. K. Michael Tse at the City University of Hong Kong. This project utilizes ordinary differential equations (ODEs) and physics-informed neural networks (PINNs) to model business growth using time series data. Work is presented in the seminar "Using AI to Model Business Growth and Its Implications for Business".
Research Interests
- Deep Reinforcement Learning (DRL)
- Artificial Intelligence (AI) for Motion Control
- Scientific AI Applications
- Reasoning Capabilities of AI (AGI)
Work Experiences
Machine Learning Intern (MAY-JULY 2023)
from: 2023, until: 2023Organization:Hong Kong Gemological Research And Authentication Centre Company Limited (Chow Sang Sang)Location:Hong Kong
Description:Segmentation and classification models, based on UNets, are utilized for diamond grading.
Software Developer, Patent Inventor (2021-2023)
from: 2021, until: 2023Organization:Centre for Intelligent Multidimensional Data Analysis Limited (CIMDA)Location:Hong Kong Science Park, Hong Kong
Description:One of the inventors of the patent AN INTERACTIVE ELDERLY ASSISTANCE SYSTEM (Patent No. HK30085199). Work includes animations using Unreal Engine (C++), Metaverse using Vircadia (Javascript), Natural Language Processing and AI Music Generation (Python), Mobile Application (Flutter&Dart).
Education
BEng, Major in Computer and Data Engineering
from: 2021, until: presentField of study:Electrical EngineeringSchool:City University of Hong KongLocation:Hong Kong