[FEATURE] Featrue Generation Pipeline

[GongJr0]

Feature Details

A pipeline module to handle the generation of numeric features, process embeddings, and format outputs into a PyTorch legible format.

Affected Modules

Exposed API regarding feature generation will be implemented in the FeatureGen class. Current modules will not recieve exposed updates.

Implementation Checklist

• Lag formatting for $AR(n)$ features
• Categorical embeddings for ticker, period, sector, etc. information
• Concatenation layer for per-ticker data
• Stateful feature storage and iterative update logic for incremental learning
• Validation utilities for shape/mask consistency
• Collate function for batching variable-length lag sets

Limitations

• The implementation will focus on pre-planned features and abstractions will only be made to facilitate efficient development of the current architecture.
• The pipeline will offer limited human-readable inspection/validation points. Format will strictly focus on torch readability.