Developing neural networks in the application of music related research, specialized in model combination, such as CNN based joint-training model and generative adversarial networks.
Data mining and symbolic feature design, aim to conduct more efficient experiments in tasks includes hit-song prediction, music recommendation, and music generation.
Also a former audio engineer, before MIR, audio circuits and equipments was my love. Starting from record demos by some simple gear that I could reach in college Rock’n Roll club, I've experienced working in local studios and touring with indie bands, the career gave me some precious memories.
While building up solid foundation in NCTU EE, my passion toward music prompted me to set the life goal of pursuing a career in which I could immerse in the beauty of music while integrating my specialty in electronic engineering
I had been working in the audio industry since my junior year in college, which made me able to gain extraordinary experiences such as being in charge of live audio mixing at a concert with thousands of audience members, or even social movements that rolled over months.
A part of my crazy college life. Somehow, my interest toward motorcycle gone too far. Which made me became a racer in a racing team and won some trophies.
Thanks to Dr. Yi-Hsuan Yang, who offered me to work as a research assistant in the Music and Audio Computing Lab at Academia Sinica in March 2016. Since then, I have been focusing on developing neural networks in the application of music related tasks, which includes hit-song prediction, music recommendation, and music generation.
After Academia Sinica, the enthusiasm for the future of AI music motivates me to pursue a master degree in GTCMT. My research covers around evaluation method of music generation, artificial intelligence in music visualization, painting creation via robotic musicianship and deep learning powered synthesizer.
Currently working with the Machine Learning Team of Bose Central Data Organization, bringing new concepts for deep learning based headphone engagement.