axect

View JSON Representation

Ph.D student of particle physics & Rustacean

Funding Links: https://github.com/sponsors/Axect

• Name: Tae-Geun Kim
• Location: Seoul, South Korea
• Company: Yonsei Univ.
• Kind: user
• Followers: 116
• Following: 75
• Repositories count: 21
• Created at: 2022-11-12T22:47:31.915Z
• Updated at: 2025-03-28T19:43:14.711Z
• Last synced at: 2025-03-28T19:43:14.710Z

GitHub Sponsors Profile

Hello, I'm Tae-Geun Kim 👋
🙋‍♂️ About Me

Graduate student in Physics at Yonsei University
Member of Yonsei HEP-COSMO
Curriculum Vitae | Blog

❤️ Interests

High energy astrophysics, dark matter, and cosmology
Scientific computation
Machine Learning / Deep Learning / Statistics
Quantum Computing

💼 Key Projects

Peroxide: Rust numeric library for scientific computing

Linear algebra, numerical analysis, statistics, and machine learning
User-friendly syntax similar to R, NumPy, and MATLAB
Supports functional programming and automatic differentiation

HyperbolicLR: Novel learning rate schedulers for deep learning

Addresses learning curve decoupling problem
Improves performance and stability across increasing epochs
Implemented and evaluated using PyTorch

Puruspe: Pure Rust library for special functions

Implements gamma, beta, and error functions with no dependencies
Lightweight and efficient for mathematical computing

Forger: Reinforcement Learning library in Rust

Modular design for agents, environments, and policies
Supports customizable strategies and learning algorithms

PyTorch Template Project: Flexible template for ML experiments

Configurable experiments using YAML files
Integration with Weights & Biases and Optuna
Support for multiple random seeds and device selection

DeeLeMa: Deep learning for particle collision analysis

Estimates mass and momenta in high-energy collider events
Adaptable to different event topologies

📚 Selected Publications

T.-G. Kim, "HyperbolicLR: Epoch insensitive learning rate scheduler", arXiv:2407.15200 (2024)
C.M. Hyun, T.-G. Kim, K. Lee, "Unsupervised sequence-to-sequence learning for automatic signal quality assessment...", CMPB 108079 (2023)
K. Ban et al., "DeeLeMa: Missing information search with Deep Learning for Mass estimation", Phys. Rev. Research 5, 043186 (2022)

Your support will help me continue developing open-source scientific computing tools and pursuing research in physics and machine learning. Thank you for considering sponsorship!

• Current Sponsors: 0
• Past Sponsors: 0
• Total Sponsors: 0
• Minimum Sponsorship: $1.00

Featured Works

Active Sponsors

Past Sponsors

Sponsor Breakdown

• User: 1