Skip to content

v0.1.0 - Public research baseline for ML-enhanced multi-factor trading

Latest
Latest

Choose a tag to compare

View all tags
@initial-d initial-d released this 05 Jul 01:18
v0.1.0
c07c295

ml-quant-trading is now ready as a public research baseline for people interested in
factor research, tensorized financial data, portfolio optimization, and reproducible
backtesting.

What is included

  • 213 factor dimensions.
  • Mask-aware PyTorch tensor factor primitives.
  • A-share oriented bias correction for limit-up, limit-down, and halt cases.
  • MLP and Transformer baselines.
  • AdjMSE, IC, and RankIC losses.
  • Cross-sectional Markowitz optimization.
  • Vectorized backtesting and metrics.
  • Synthetic-data end-to-end pipeline.
  • Public-data factor IC notebook.
  • CPU/GPU tensor factor benchmark script.
  • CI, tests, citation metadata, and contribution templates.

Quick start

git clone https://github.com/initial-d/ml-quant-trading.git
cd ml-quant-trading
pip install -e .[dev]
make paper CONFIG=configs/small.yaml

Community ask

I would especially appreciate:

  • benchmark results from different CPUs/GPUs
  • public-data reproduction reports
  • factor-engine edge cases
  • documentation fixes
  • small examples that make the project easier to learn

Use the benchmark issue template to submit performance results:
#12

For open-ended reports and comparisons, use the benchmark/reproduction discussion:
#13

Good first contribution ideas:

  • more CPU/GPU benchmark reports: #7
  • ETF or larger-universe public-data reproduction: #16

The benchmark board now includes the first maintainer CPU baseline:
https://github.com/initial-d/ml-quant-trading/blob/main/docs/benchmark_board.md

The docs now include a public-data mini reproduction on yfinance data:
https://github.com/initial-d/ml-quant-trading/blob/main/docs/public_data_mini_reproduction.md

The docs also include a larger public-data validation benchmark with a 100-stock
yfinance reference run, walk-forward baselines, transaction costs, slippage,
turnover, drawdown, and equal-weight / momentum / Alpha101 / MLP / Transformer
comparisons:
https://github.com/initial-d/ml-quant-trading/blob/main/docs/public_data_validation.md

The validation boundary is documented explicitly here:
https://github.com/initial-d/ml-quant-trading/blob/main/docs/reality_check.md

Recent contributor-friendly updates:

  • Start Here now includes expected runtime, hardware requirements, artifact paths, and typical successful output.
  • A VS Code / Codespaces Dev Container is included for reproducible setup.
  • The architecture page now shows the data -> factor -> model -> portfolio -> backtest pipeline.
  • Benchmark reporting now asks for CUDA version in addition to CPU, GPU, PyTorch, and OS.

The social preview image is attached to this release for launch posts and community sharing.

Disclaimer

This repository is for research and engineering experimentation. It is not financial
advice, investment advice, or a trading recommendation.

Assets 3
Loading