Thang D. Chu

Hello world! Welcome to my personal webpage.

I am 2nd year M.Sc. student @ University of Alberta. I am fortunately supervised by professor Csaba Szepesvari and a member of RLAI lab. I am a ML theory Community lead, used to be a RL Community Lead @ CohereAI

I am interested in understanding and developing algorithms rigorously. A current list of my interests:

RL/ML theory
Online Learning / Optimization

news

Sep 1, 2024	I will be TA again for CMPUT 267: Machine Learning 1.
Jun 30, 2024	I will attend and be a volunteer for COLT 2024. Very excited to learn more about learning theory!!
Mar 18, 2024	I am fortunately supervised by professor Csaba Szepesvári. My goal is to have a better theoretical understanding about policy gradient and actor-critic methods!
Jan 8, 2024	I will be a TA for CMPUT 267: Basics of Machine Learning taught by Nidhi Hedge.
May 28, 2023	I will join the Department of Computing Science at University of Alberta to pursue my Msc degree in Winter 2024.

latest posts

Feb 1, 2022	a post with redirect

selected publications

IEEE/ACM

Graph Transformer for drug response prediction

Thang Chu, Thuy Trang Nguyen, Bui Duong Hai, and 2 more authors

IEEE/ACM Transactions on Computational Biology and Bioinformatics, Sep 2022

Abs HTML PDF

Previous models have shown that learning drug features from their graph representation is more efficient than learning from their strings or numeric representations. Furthermore, integrating multi-omics data of cell lines increases the performance of drug response prediction. However, these models have shown drawbacks in extracting drug features from graph representation and incorporating redundancy information from multi-omics data. This paper proposes a deep learning model, GraTransDRP, to better drug representation and reduce information redundancy. First, the Graph transformer was utilized to extract the drug representation more efficiently. Next, Convolutional neural networks were used to learn the mutation, meth, and transcriptomics features. However, the dimension of transcriptomics features was up to 17737. Therefore, KernelPCA was applied to transcriptomics features to reduce the dimension and transform them into a dense presentation before putting them through the CNN model. Finally, drug and omics features were combined to predict a response value by a fully connected network. Experimental results show that our model outperforms some state-of-the-art methods, including GraphDRP and GraOmicDRP. Availability of data and materials : https://github.com/chuducthang77/GraTransDRP.