Hi, I'm Arnold.
I do machine learning and data science.
I am an Applied Scientist at Amazon. I have also been in R&D roles at TikTok, ASMPT, and Scotiabank.
I completed my MSc in Computer Science at the University of Toronto (2018 - 2020), while conducting machine learning research at the Vector Institute for AI.
Contact me directly at arnold [dot] ys [dot] yeung [at] gmail [dot] com
Select Machine Learning Projects
Personalized Self-Explaining Image Classifier
Developed an RL framework which provides end users with sequential personalized explanations for the behavior of an image classifier.
The AI agent will predict what explanations you'll find most helpful as you learn how to simulate its behavior.
Sponsored by the Electronics and Telecommunications Institute (ETRI) of the Republic of Korea.
Published and presented at ICML 2020 Workshop on Human Interpretability in Machine Learning (Spotlight).
Machine Learning Prediction of Confirmed COVID-19 Cases using Non-Pharmaceutical Interventions and Cultural Dimensions
Developed an ETL pipeline for 3 datasets, containing non-pharmaceutical interventions and cultural dimensions of 114 countries.
Trained multiple machine learning models to predict the confirmed infection growth in the following 14 days. Evaluated model performance using 3 validation methods, each applicable to a different use case.
Published in Journal of Medical Internet Research, 2021.
Detecting Phonemic Confusion using an LSTM
Developed a bidirectional LSTM network with skip connections to identify phonemes in text which may be misheard by a proxy listener.
Compared with a feed-forward fully-connected network and a Maximum Likelihood Estimate baseline. Evaluated using a Levenshtein distance metric.
Published and presented at ACL-IJCNLP 2021 SIGMORPHON.
Battery State-of-Health Determination
Developed a machine learning model which determines the state-of-health of mobile phone batteries using the Cadex C5100 Battery Analyzer.
Completed at and patented by Cadex Electronics, 2015.
[patent]
Image Colorization with Convolutional Neural Networks
Implemented and trained two CNN architectures (vanilla and U-Net) to colorize greyscale images from the CIFAR-10 dataset.
[code]
Select Engineering Projects
Comparing the Performances of Wet and Dry EEG Electrodes
Designed and constructed an EEG headcap with dry and wet electrodes and conducted human experimentation. Evaluated and compared the performances of different dry electrodes, relative to standard wet electrodes.
Sponsored by the Natural Sciences and Engineering Research Council (NSERC) of Canada.
Published and presented at IEEE Engineering in Medicine and Biology Conference (EMBC) 2015 (Oral Presentation).
[paper]
Retail Teabag Packager
Identified critical needs of small-scale specialty tea-leaves retailers. Conceptualized and designed an electro-mechanical teabag packager. Manufactured and assembled a physical MVP prototype for demonstration purposes.
[slides]
Concussion Risk Assessment Tool
Developed a framework for assessing the risk of concussion in a sports team based environment.
Sponsored by the Natural Sciences and Engineering Research Council (NSERC) of Canada.
Published in Brain Injury, 2017.
[paper]