Skip to content

roadmap.md

Latest commit

 

History

History
61 lines (48 loc) · 2.76 KB

roadmap.md

File metadata and controls

61 lines (48 loc) · 2.76 KB
Raw

Roadmap

This roadmap is organized around contributions that make the project easier to run, easier to trust, and easier to extend.

Near Term

  • Collect larger public-data validation reports using scripts/public_data_validation.py.
  • Collect benchmark results from different CPUs, GPUs, CUDA versions, and PyTorch versions.
  • Expand first-run onboarding based on new user feedback.
  • Add more examples for factor IC, attribution, and regime-specific diagnostics.
  • Add independent reproductions with explicit survivorship-bias and point-in-time data controls.
  • Evaluate a lightweight GitHub Project board for contributor tasks.
  • Evaluate automated pull-request review tooling after the first external PRs arrive.

Completed Launch Items

  • Published v0.1.0 as the first public research baseline.
  • Added a maintainer CPU benchmark baseline to docs/benchmark_board.md.
  • Added a yfinance public-data mini reproduction to docs/public_data_mini_reproduction.md.
  • Added a larger public-data validation harness with walk-forward baselines, costs, slippage, turnover, and drawdown.
  • Added docs/reality_check.md to separate engineering validation, smoke tests, data-gated paper reproduction, and production-readiness claims.
  • Added a Dev Container for reproducible contributor setup.
  • Added a root Dockerfile and Docker setup guide for CPU reproduction.
  • Added a Mermaid architecture diagram in docs/architecture.md.

Contributor-Friendly Tasks

  • Translate key docs between English and Chinese.
  • Add tests for neutralization and backtest edge cases.
  • Add a small example using a custom CSV data source.
  • Add benchmark results through the benchmark issue template.
  • Run scripts/public_data_validation.py on a new public-data universe and report the exact command.
  • Improve docstrings for factor families.
  • Add one new ETF or larger-universe public-data example with a clearly documented universe.

Research Extensions

  • Expand Alpha101 formula coverage.
  • Add factor selection examples.
  • Add cross-validation and walk-forward evaluation examples.
  • Add ablation scripts for bias correction, losses, and transaction costs.
  • Add portfolio attribution reports.

Engineering Extensions

  • Add optional GPU benchmark reporting.
  • Add parquet-based data loading examples.
  • Add reproducible environment files for CUDA and CPU-only users.
  • Add a CUDA-specific Dockerfile or documented NVIDIA container workflow.
  • Add a minimal web dashboard for benchmark and backtest summaries.
  • Tighten linting gradually after a dedicated formatting pass.

Community Milestones

  • First external benchmark result.
  • First public-data reproduction issue.
  • First external PR.
  • First tagged release.
  • First Zenodo archive or DOI-backed software release.
  • First third-party tutorial or blog post.