Research interets
- Mathematics
- Geometric analysis on graphs
- Spectral geometry
- MicroeconomyTheory
- Mechanism design
- Auction theory
- Market Segmentation
- Computer Science
- Machine Learning Theory （theortical interpretability）
- prompt learning and in-context learning
- Dialogue Systems
- Machine Learning’s application in mathematics and economics（Algorithm mechanism design）
Selected Projects
1. Asset Pricing via Machine learning
Supervisor: Prof. Lei Ge
Mar. 2023 — Jun. 2023, RUC
In this project, I implemented machine learning model (including OLS ,Lasso, RF, XGboost, Lgbm, NN3), to forecast stock reutrns and identify potential market trends.
2.Note on Geometric Analysis on Graph
Jul.2023 – Aug.2023
A personal latex note in the field of geometric analysis on the graph,which covers following three topics: spectral graph theory, partial differential equations on graphs and Ricci curvature on graphs.
3. ATRI-Reproduce
Aug.2023 – (ongoing)
In this project, I have attempted to mimic the distinctive speech patterns of a particular anime character by fine-tuning the large language model ChatGLM-6B. Capturing the stylistic qualities of fictional personas presents an intriguing challenge for training AI systems, requiring models to learn more complex representations of linguistic styles.
At present, the model has developed robust general linguistic capabilities by p-training on datasets.However, I soon realized this project might be better suited for a generative model, since I intend to make the LLM speak like the character instead of speaking exactly the same as the character. The key focus in overcoming this challenge is designing an appropriate loss function that encapsulates the defining features of the character’s dialogue. This may include vocabulary, grammar, tone, sentence structure and other stylistic elements. The loss function can penalize deviations from these patterns while rewarding mimicking them.