Skip to content

NTK-Change-Point-Detection

Neural Tangent Kernel Dynamics for Change Point Detection in Financial Time Series

View on GitHub Homepage

Information

Property Value
Language CSS
Stars 0
Forks 0
Watchers 0
Open Issues 0
License MIT License
Created 2026-01-19
Last Updated 2026-02-19
Last Push 2026-01-19
Contributors 1
Default Branch main
Visibility private

Datasets

This repository includes 3 dataset(s):

Dataset Format Size

| data | | 0.0 KB |

| algorithms.json | .json | 6.12 KB |

| equations.json | .json | 19.76 KB |

Reproducibility

No specific reproducibility files found.

Status

  • Issues: Enabled
  • Wiki: Enabled
  • Pages: Disabled

README

NTK Change Point Detection

GitHub Pages License: MIT

Neural Tangent Kernel Dynamics for Change Point Detection in Financial Time Series

Overview

This repository contains the research website and supplementary materials for our paper on using Neural Tangent Kernel (NTK) theory for detecting structural changes in financial time series.

Live Site: https://digital-ai-finance.github.io/NTK-Change-Point-Detection/

Authors

  • Ruochen Wu - Humboldt University Berlin
  • Xiaorui Zuo - Humboldt University Berlin
  • Liangliang Zhang - Humboldt University Berlin
  • Wolfgang Karl Hardle - Humboldt University Berlin, Singapore Management University
  • ORCID: 0000-0001-5600-3014

Key Contributions

  1. NTK-based Change Point Detection: A novel framework leveraging empirical NTK computation for detecting regime changes
  2. Rolling Window Segmentation: BIC-penalized SSE minimization with tolerance-based clustering
  3. Walk-Forward Training: Change-point-aware training with multiscale patches and controlled forgetting
  4. Portfolio Application: Variable-window Transformer for adaptive mean-variance optimization

Repository Structure

docs/
  index.html           # Main single-page application
  css/
    main.css           # Core layout and design system
    components.css     # UI components (cards, buttons, etc.)
    visualizations.css # Plotly.js chart styling
    equations.css      # MathJax and algorithm blocks
  js/
    main.js            # App initialization and utilities
    visualizations.js  # 8 Plotly.js interactive charts
    openalex.js        # OpenAlex API for publications
    bibtex.js          # Citation widget
  data/
    equations.json     # 64 LaTeX equations
    algorithms.json    # 2 algorithm pseudocodes
  assets/
    images/            # Logos and images

Technologies

  • MathJax 3.x - LaTeX equation rendering
  • Plotly.js 2.27 - Interactive visualizations
  • Inter Font - Typography
  • OpenAlex API - Publication metadata
  • GitHub Pages - Static hosting

Local Development

# Clone the repository
git clone https://github.com/Digital-AI-Finance/NTK-Change-Point-Detection.git
cd NTK-Change-Point-Detection

# Serve locally (requires Python)
python -m http.server 8000

# Open in browser
open http://localhost:8000

Citation

@article{wu2025ntk,
  title={Neural Tangent Kernel Dynamics for Change Point Detection
         in Financial Time Series},
  author={Wu, Ruochen and Zuo, Xiaorui and Zhang, Liangliang
          and H{\"a}rdle, Wolfgang Karl},
  journal={Working Paper},
  year={2025},
  url={https://digital-ai-finance.github.io/NTK-Change-Point-Detection/}
}

License

MIT License - see LICENSE for details.

Organization

Part of the Digital-AI-Finance research initiative.