A dedicated graduate student with a strong foundation in technology and research, specializing in groundbreaking projects ranging from machine leaning to neuromorphic engineering. Demonstrated expertise in data processing and deep learning, proficient in applying innovative solutions to complex problems, leveraging interdisciplinary knowledge to drive progress.
Generating Voices with Frontier TTS Technology:
This project explored the performance of a state-of-the-art end-to-end Text-to-Speech (TTS) model, specifically focusing on variational inference and adversarial learning processes. We applied data preprocessing techniques and hyperparameter tuning to pre-trained VITS models, adapting them to our professor's voice. The experiment with the VCTK pre-trained model demonstrated significant achievements in voice generation with limited resources.
Enhancing Precipitation Forecasting with Hybrid NWP-DL Models:
In this innovative project, we developed a hybrid model combining Numerical Weather Prediction (NWP) and Deep Learning (DL) for improved precipitation forecasting. By integrating the strengths of traditional NWP models with advanced DL techniques, our model efficiently processed the KoMet dataset, featuring complex atmospheric data from the Korean Peninsula. Utilizing various RNN architectures and optimizing parameters like window size and lead time, we achieved notable improvements in forecast accuracy.
Advancements in Synaptic Device Technology (Undergraduate Final Year Project)
This project entailed the development and testing of synaptic devices, focusing on the preparation of silicon wafers, application of semiconductor layers through spin-coating, and electrode plating using aluminum. I played a key role in the experimental setup, including processing and testing devices under controlled conditions to analyze the current-voltage relationship. Additionally, I integrated these devices into a Spiking Neural Network (SNN), utilizing Python for applying device parameters and employing the Spike-Timing-Dependent Plasticity (STDP) method for network training, contributing to the field of neuromorphic engineering.
Head of Project Development Department of Science and Technology Innovation Association of XJTLU, I Taught members how to use SolidWorks for 3D modeling and organized 3D-printing Exhibition, a total of six or seven groups of works were displayed, and nearly 100 people visited.