CV

Education

• B.S. in Mathematics and Applied Mathematics, Xi’an Jiaotong University, Sep 2024–present

Research Experience

Low-Frequency Spectral Bias of Neural Networks in Time-Series Forecasting
Sep 2024–Feb 2025 | Adviser: Prof. Xuan Sun, Xi’an Jiaotong University

• Provided the first mathematical proof of the “frequency principle” phenomenon in neural networks.
• Introduced FreLE, an algorithm that explicitly/implicitly regularizes frequency components; achieves SOTA on ETT & Weather datasets.
• Paper has been accepted as a short paper at IEEE ICDM 2025.

Deep-Learning-Based Seismic Horizon Picking
Feb 2025–Apr 2025 | Adviser: Prof. Naihao Liu, Xi’an Jiaotong University

• Designed Channel-Independent Multi-Scale UNet (CIMS-UNet): multi-period embeddings, multi-scale kernels, channel-independent training strategy.
• 2 papers (1st-author & co-author) under review.

Deep Learning for 3D Molecular Structure Reconstruction from EM Images
Feb 2025– Aug 2025 | Collaborating with Prof. Lei Zhang, Xi’an Jiaotong University

• In charge of denoising: BM3D, improved Noise2Noise & Noise2Self pipelines; boosts downstream 3D reconstruction quality.

Publications

1. Sun, J., Ling, X., Zou, J., Kang, J., & Zhang, K. (2025). FreLE: Low-Frequency Spectral Bias in Neural Networks for Time-Series Tasks. Proceedings of the IEEE International Conference on Data Mining (ICDM 2025). (Accepted)

2. Zou, J., Liu, N., Zhang, T., He, J., Li, T., & Gao, J. (2025). Seismic Horizon Picking Using Channel-Independent Multi-Scale Network with Limited Training Samples. Geophysics. (Under Review)

3. Zhang, T., He, J., Liu, N., Zou, J., & Jiang, Y. (2025). Mamba-Driven and Feature-Fused U-Net for Automatic Seismic Horizon Interpretation. IEEE Transactions on Geoscience and Remote Sensing (TGRS). (Under Review)

Research Interests

Deep-learning theory, large-model mechanisms, natural-language processing, AI4Science.

Outreach

• Deep Learning Seminar (founder & lecturer)
  Sep 2024–present, nationwide online seminar originating from Xi’an Jiaotong University; 900+ participants from Tsinghua, PKU, CAS, etc.

Future Plan

Pursue theoretically-oriented machine learning research, especially the internal mechanisms of large language models—representation, training dynamics, and generalization—using rigorous mathematics to build interpretable and trustworthy AI.

Skills

• Programming: Python, C++
• Frameworks/Libraries: PyTorch, NumPy, Transformers, LaTeX