Currently Assistant Professor at Tokyo Institute of Technology since April, 2022.

My main research focus is on bridging the gap between symbolic AI and neural networks.

During my Ph.D., I focused on developing deep learning techniques that could extract symbolic rules from time series data. The symbolic rules generated by these techniques were meant to serve as an explanation for the observed data, and could be used for prediction and manipulation of dynamic systems. My work aimed to create explainable AI systems that could provide human-accessible insights into complex systems.

Currently, I am exploring a new technique that combines the strengths of both symbolic and neural networks. While symbolic AI provides insights that are directly transferable and learnable by humans, its brittleness limits its real-world application. On the other hand, neural networks are robust against noise and have broad real-world applications, but the knowledge and insights they provide can be difficult to access. By combining both approaches, I hope to devise intelligent systems that are both robust and accessible to humans. My goal is to create explainable AI systems that can provide valuable insights into complex systems while also being practical for real-world applications.

Research areas: deep learning, symbolic rule extraction, time series analysis, explainable AI, hybrid approach, neural networks, robustness, and human-accessible insights.

Have interesting projects or want to have some discussion? You can contact me at phua [at] c.titech.ac.jp

Goal

My main focus is on bridging the gap between symbolic AI and neural networks, combining the strengths of both approaches to create more robust and flexible intelligent systems.

Main Publications

Class-Incremental Learning using Diffusion Model for Distillation and Replay [Best Paper Award]
Quentin Jodelet, Xin Liu, Yin Jun Phua, Tsuyoshi Murata
1st Workshop on Visual Continual Learning, ICCV2023, 2023.

resVAE ensemble: Unsupervised identification of gene sets in multi-modal single-cell sequencing data using deep ensembles
Foo Wei Ten, Dong Sheng Yuan, Nabil Jabareen, Yin Jun Phua, Richard Eils, Soeren Lukassen, Christian Conrad
Frontiers in Cell and Developmental Biology 11, 2023.

Learning Logic Programs Using Neural Networks by Exploiting Symbolic Invariance
Yin Jun Phua, Katsumi Inoue
ILP, 203-218, 2021.

Learning Logic Programs from Noisy State Transition Data
Yin Jun Phua, Katsumi Inoue
Inductive Logic Programming - 29th International Conference(ILP), JSAI2019 72-80, 2019.

Learning Representation of Relational Dynamics with Delays and Refining with Prior Knowledge [Best Student Paper Award]
Yin Jun Phua, Tony Ribeiro, Katsumi Inoue
IfCoLog Journal of Logics and their Applications(FLAP), 6(4) 695-708, 2019.

Learning Logic Program Representation for Delayed Systems With Limited Training Data
Yin Jun Phua, Tony Ribeiro, Sophie Tourret, Katsumi Inoue
Late Breaking Papers of the 27th International Conference on Inductive Logic Programming, 27-37, 2017.

Research Projects/Grants

Research on Intelligence and Consciousness
Principal Investigator ￥ 500,000
New Research Challenge Incentive Fund,
Organization for Fundamental Research, Tokyo Institute of Technology
September 2023 - March 2024

Discovering New Knowledge by Combining Symbolic Logic and Deep Learning
Principal Investigator ￥ 2,860,000
Grants-in-Aid for Scientific Research Grant-in-Aid for Research Activity Start-up,
Japan Society for the Promotion of Science
August 2022 - March 2024

Robust AI by Integration of Knowledge Representation and Machine Learning
Co-Investigator ￥ 41,470,000
Grant-in-Aid for Scientific Research (A),
Japan Society for the Promotion of Science
April 2022 - March 2025

Understandable Model by Combination of Symbolic Machine Learning and Deep Statistical Learning
Research Fellow ￥ 1,500,000
Grants-in-Aid for Scientific Research Grant-in-Aid for JSPS Fellows,
Japan Society for the Promotion of Science
April 2021 - March 2022