VeighNa 4.0 Adds a Machine Learning Toolkit for Quantitative Strategies

Summary

This release overview describes VeighNa 4.0 and its new vnpy.alpha module for developing machine-learning, multi-factor strategies. The module is organized around feature datasets, model training, strategy research, workflow management, and example notebooks. It lists Alpha 158 features and model options including Lasso, LightGBM, and multilayer perceptrons, and says the framework supports both cross-sectional multi-asset strategies and time-series single-asset strategies. The described workflow includes data management, signal generation, and backtesting.

The article also reports a move to Python 3.13 and changes to the project's build, type checking, linting, development environment, and logging tools. It notes that some extensions were adapted for the new release while other module updates were planned for a later version. This is a software capability announcement, not an empirical strategy study: it provides no performance results, validation details, or guidance on avoiding common machine-learning pitfalls such as leakage, overfitting, and survivorship bias.

Key ideas

• VeighNa 4.0 introduces vnpy.alpha for machine-learning strategy research and trading workflows.
• The module supports feature engineering, model training, signal generation, and backtesting.
• Listed model choices include Lasso, LightGBM, and multilayer perceptrons.
• The examples cover both cross-sectional multi-asset and single-asset time-series approaches.
• The release announcement provides no evidence of strategy profitability or model robustness.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.