Crate linfa_clustering

linfa-clustering aims to provide pure Rust implementations of popular clustering algorithms.

The big picture

linfa-clustering is a crate in the linfa ecosystem, a wider effort to bootstrap a toolkit for classical Machine Learning implemented in pure Rust, kin in spirit to Python’s scikit-learn.

You can find a roadmap (and a selection of good first issues) here - contributors are more than welcome!

Current state

Right now linfa-clustering provides the following clustering algorithms:

• K-Means
• DBSCAN
• Approximated DBSCAN
• Gaussian-Mixture-Model

Implementation choices, algorithmic details and tutorials can be found in the page dedicated to the specific algorithms.

Additionally, this crate provides the generate_blobs utility to quickly generate test datasets for clustering.

Check here for extensive benchmarks against scikit-learn’s K-means implementation.

Structs

AppxDbscan	DBSCAN (Density-based Spatial Clustering of Applications with Noise) clusters together neighbouring points, while points in sparse regions are labelled as noise. Since points may be part of a cluster or noise the transform method returns Array1<Option<usize>>. It should be noted that some “border” points may technically belong to more than one cluster but, since the transform function returns only one label per point (if any), then only one cluster is chosen arbitrarily for those points.
AppxDbscanHyperParams	The set of hyperparameters that can be specified for the execution of the Approximated DBSCAN algorithm.
AppxDbscanHyperParamsBuilder	Helper struct used to construct a set of hyperparameters for the approximated DBSCAN algorithm
AppxDbscanLabeler	Struct that labels a set of points according to the Approximated DBSCAN algorithm
Dbscan	DBSCAN (Density-based Spatial Clustering of Applications with Noise) clusters together points which are close together with enough neighbors labelled points which are sparsely neighbored as noise. As points may be part of a cluster or noise the predict method returns Array1<Option<usize>>
DbscanHyperParams	The set of hyperparameters that can be specified for the execution of the DBSCAN algorithm.
DbscanHyperParamsBuilder	Helper struct used to construct a set of hyperparameters for DBSCAN algorithm.
GaussianMixtureModel	Gaussian Mixture Model (GMM) aims at clustering a dataset by finding normally distributed sub datasets (hence the Gaussian Mixture name) .
GmmHyperParams	The set of hyperparameters that can be specified for the execution of the GMM algorithm.
KMeans	K-means clustering aims to partition a set of unlabeled observations into clusters, where each observation belongs to the cluster with the nearest mean.
KMeansHyperParams	The set of hyperparameters that can be specified for the execution of the K-means algorithm.
KMeansHyperParamsBuilder	An helper struct used to construct a set of valid hyperparameters for the K-means algorithm (using the builder pattern).

Enums

GmmCovarType	A specifier for the type of the relation between components’ covariances.
GmmError	An error when modeling a GMM algorithm
GmmInitMethod	A specifier for the method used for the initialization of the fitting algorithm of GMM
KMeansError	An error when modeling a KMeans algorithm
KMeansInit	Specifies centroid initialization algorithm for KMeans.

Functions

compute_inertia	We compute inertia defined as the sum of the squared distances of the closest centroid for all observations.
generate_blob	Generate blob_size data points (a “blob”) around blob_centroid.
generate_blobs	Given an input matrix blob_centroids, with shape (n_blobs, n_features), generate blob_size data points (a “blob”) around each of the blob centroids.