rust-ml/linfa

Features

• Classical Machine Learning - Implements a comprehensive toolkit for traditional statistical learning including regression, clustering, and decision trees.
• Data Preprocessing Toolkits - Provides extensive utilities for feature normalization, text vectorization, and dimensionality reduction.
• Model Ensembling - Implements ensemble methods like bagging and boosting to combine multiple learners into a more robust model.
• Rust Machine Learning Libraries - Provides a complete collection of machine learning algorithms implemented specifically in the Rust language.
• Statistical Learning Implementations - Implements programmatic applications of statistical learning theory for fitting models to labeled and unlabeled datasets.
• Multi-Language Wasm Compilers - Compiles machine learning logic into WebAssembly binaries to enable execution within browser environments.
• Linear Algebra - Accelerates matrix and vector operations by linking to external high-performance linear algebra backends.
• Tensor Operations - Employs multidimensional arrays via ndarray for storing and manipulating datasets and feature vectors.
• Model Training Interfaces - Standardizes model training and prediction through a common trait-based interface for diverse learning algorithms.
• WebAssembly Compilation - Provides WebAssembly compilation for machine learning logic to ensure browser-based compatibility.
• Hierarchical Clustering - Groups data points into nested hierarchies by iteratively merging the closest pairs.
• Decision Trees - Implements decision tree models that use tree-like structures to map observations to target values.
• Ensemble Learning - Combines multiple algorithms into a single model using Random Forest, AdaBoost, and Bagging.
• Bootstrap Aggregation - Combines multiple base classifiers into an ensemble via bagging on random data subsets.
• Feature Scale Normalization - Prepares datasets through whitening, count vectorization, and TF-IDF to ensure stable model convergence.
• Clustering Algorithms - Implements various unsupervised grouping techniques including K-Means, DBSCAN, and Gaussian Mixture Models.
• Kernel-Based Feature Mapping - Uses RBF and polynomial kernels to map data into higher-dimensional spaces for non-linear classification.
• Linear Regression Models - Computes relationships between variables using Ordinary Least Squares or Generalized Linear Models.
• Support Vector Regression - Implements support vector regression for continuous value prediction using Epsilon and Nu-SVM settings.
• Elastic Net Regression - Fits linear models by combining L1 and L2 penalties to perform feature selection.
• Least Angle Regression - Fits linear models to data using the least-angle regression algorithm for sparse coefficients.
• Partial Least Squares Regressions - Predicts a response variable by projecting predictors and responses into a lower-dimensional space to maximize covariance.
• Logistic Regression Models - Provides algorithms for predicting binary or multinomial outcomes using logistic regression.
• Model Performance Evaluators - Ships tools for quantifying model accuracy and error using confusion matrices and cross-validation.
• Regularized Regression - Provides target regression using shrinkage operators such as LASSO and Ridge to prevent overfitting.
• Naive Bayes Classifiers - Implements probabilistic classification models based on Gaussian, Multinomial, and Bernoulli distributions.
• Text Vectorizations - Transforms raw text into numerical representations using count-based and TF-IDF vectorization.
• Supervised Classification - Provides general workflows for training models to categorize data using labeled datasets.
• Dimensionality Reduction - Reduces high-dimensional feature vectors using techniques like PCA and t-SNE to improve model efficiency.
• Support Vector Classifiers - Builds classification models that separate data using C, Nu, or one-class SVM settings.
• Training Dataset Preparation - Prepares raw data from CSV or compressed formats and transforms features for model training.
• Dimensionality Projection Plots - Projects high-dimensional data into lower-dimensional spaces using exact or Barnes-Hut t-SNE for cluster visualization.
• Numerical Scaling - Rescales numeric data using standard, min-max, or maximum absolute methods to normalize feature contribution.
• Dimensionality Reduction - Transforms feature vectors using PCA, Diffusion mapping, and t-SNE to map data into different dimensional spaces.
• Vector Scaling and Whitening - Adjusts the magnitude of data vectors and applies whitening to remove correlations and standardize variance.
• Spatial Proximity Computations - Calculates nearest neighbors in a dataset using various distance functions and spatial index structures.
• Linear Algebra Libraries - Provides a backend-agnostic layer to switch between native Rust and high-performance linear algebra libraries.
• Nearest Neighbor Searches - Accelerates nearest neighbor lookups using spatial index structures like KD-trees and ball trees.
• ML Model Compilation - Enables analytical models to execute within browser environments via WebAssembly compilation.
• Artificial Intelligence - Comprehensive machine learning framework.
• Machine Learning Frameworks - Comprehensive machine learning framework.
• Machine Learning Frameworks - Toolkit for building machine learning applications in Rust.

Star history