合抱之木，生于毫末
I am a PhD Candidate in the Analysis and Interpretation of Biomedical Data Group (AID Group) within the Department of Cybernetics at the Czech Technical University. My research focuses on Artificial Intelligence and Biocybernetics, under the supervision of doc. Ing. Daniel Novák, Ph.D. This work is generously supported by the Research Center for Informatics (a EU Project), the Brain Dynamics (a EU and Czech National Grant) initiative, and Student Grants from the Czech Technical University in Prague. My academic journey includes a strong foundation in Neuroscience, gained through studies with Prof. Yong Hu, Director of the Neural Engineering & Clinical Electrophysiology Laboratory at the Li Ka Shing Faculty of Medicine, The University of Hong Kong. I further enriched my expertise during a visit to the Knowledge Discovery and Machine Learning Lab at the University of Leicester, where I was supervised by Prof. Huiyu Zhou .
During my Doctoral study (2020-2025), I conducted in-depth research on the theoretical foundations of artificial intelligence via control theory, including the currently popular large models. This opens a new door for the development, design, evaluation, and optimization of artificial intelligence in the industry.
During my Master’s and Doctoral studies over the past 10 years (2014-2024), I developed a non-invasive brain-computer interface-based depression grading diagnosis system. This system provides with a visualization of brain networks and is embedded into LLMs to achieve multimodal interconnection. This toolkit was applied to LLMs, enhancing their multimodality performance in time-series signal analysis, particularly in medical time-series signals.
- 2025/4/5 - Optimized SGD by fuzzy logic, most artificial neural networks can be learned more stable. This paper has been accepted by Expert Systems with Applications. Get the paper and codes.
- 2024/12/12 - We optimized the hallucination of generation models by formalising the system of generation models as a class of stochastic dynamical systems through the lens of control theory. Get the paper and codes.
- 2024/11/27 - Utilizing a symbolic time series representation method, we made LLMs understand time series better . Get the paper and codes.
- 2024/10/1 - Optimizing LoRA by the Shunting Inhibition and Disinhibition Mechanism, we enhanced GLUE benchmark and SQuAD performance by approximately 5%. Get the paper and codes.
Languages: Mandarin (Native Speaker), English (Proficiency level), Cantonese (Beginner)
Interests: Backpacking, Hiking, Reading, Playing Board Games