Become a sponsor to Tae-Geun Kim
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!
1 sponsor has funded Axect’s work.
Featured work
-
Axect/Peroxide
Rust numeric library with high performance and friendly syntax
Rust 530 -
Axect/Socialst
Axect's Customization Files
TeX 7 -
Axect/puruspe
PURe RUSt SPEcial library
Rust 15 -
Axect/HNumeric
Haskell Numeric Library (Pure Functional, MATLAB & R Syntax)
Haskell 8 -
Axect/Peroxide_Gallery
Examples of Peroxide (Rust numeric library)
Rust 11 -
Axect/NCDataFrame.jl
Read & write netcdf file via DataFrames
Julia 5