Expand description
Sparse feature representation for high-dimensional data.
RillML uses a sorted (FeatureId, value) vector instead of
HashMap<String, f64> as the core sparse representation. This keeps
serialization deterministic and memory predictable.
Structs§
- Sparse
Features - Sparse feature vector: sorted
(FeatureId, value)pairs.
Type Aliases§
- Feature
Id - Feature identifier type. Use integers, not strings.