incremental-rs
Out-of-Core and Incremental Model Fitting for Rust.
Overview
incremental-rs fills the gap in the Rust ML ecosystem where estimators expect fully in-memory dataset arrays. It provides partial_fit implementations matching scikit-learn's incremental capability scope.
Estimator Capabilities
| Estimator | Incremental Algorithm | Trait |
|---|---|---|
IncrementalLinearRegression |
Stochastic Gradient Descent (MSE) | IncrementalSupervisedEstimator |
IncrementalLogisticRegression |
SGD / Cross-Entropy / OvR | IncrementalSupervisedEstimator |
IncrementalGaussianNaiveBayes |
Welford's Online Means/Variance | IncrementalSupervisedEstimator |
MiniBatchKMeans |
Sculley (2010) Web-Scale K-Means | IncrementalUnsupervisedEstimator |
Out-of-Scope Design
Decision trees, Random Forests, and exact closed-form Ordinary Least Squares (OLS) do not have direct incremental variants and are intentionally excluded from online fitting scope.
##Licensed under MIT.