Merge pull request #44 from smartcorelib/api

feat: consolidates API

Author: VolodymyrOrlov

src/api.rs

@@ -0,0 +1,43 @@
+//! # Common Interfaces and API
+//!
+//! This module provides interfaces and uniform API with simple conventions
+//! that are used in other modules for supervised and unsupervised learning.
+
+use crate::error::Failed;
+
+/// An estimator for unsupervised learning, that provides method `fit` to learn from data
+pub trait UnsupervisedEstimator<X, P> {
+    /// Fit a model to a training dataset, estimate model's parameters.
+    /// * `x` - _NxM_ matrix with _N_ observations and _M_ features in each observation.
+    /// * `parameters` - hyperparameters of an algorithm
+    fn fit(x: &X, parameters: P) -> Result<Self, Failed>
+    where
+        Self: Sized,
+        P: Clone;
+}
+
+/// An estimator for supervised learning, , that provides method `fit` to learn from data and training values
+pub trait SupervisedEstimator<X, Y, P> {
+    /// Fit a model to a training dataset, estimate model's parameters.
+    /// * `x` - _NxM_ matrix with _N_ observations and _M_ features in each observation.
+    /// * `y` - target training values of size _N_.
+    /// * `parameters` - hyperparameters of an algorithm
+    fn fit(x: &X, y: &Y, parameters: P) -> Result<Self, Failed>
+    where
+        Self: Sized,
+        P: Clone;
+}
+
+/// Implements method predict that estimates target value from new data
+pub trait Predictor<X, Y> {
+    /// Estimate target values from new data.
+    /// * `x` - _NxM_ matrix with _N_ observations and _M_ features in each observation.
+    fn predict(&self, x: &X) -> Result<Y, Failed>;
+}
+
+/// Implements method transform that filters or modifies input data
+pub trait Transformer<X> {
+    /// Transform data by modifying or filtering it
+    /// * `x` - _NxM_ matrix with _N_ observations and _M_ features in each observation.
+    fn transform(&self, x: &X) -> Result<X, Failed>;
+}

src/base.rs

@@ -15,8 +15,7 @@
 //! let blobs = generator::make_blobs(100, 2, 3);
 //! let x = DenseMatrix::from_vec(blobs.num_samples, blobs.num_features, &blobs.data);
 //! // Fit the algorithm and predict cluster labels
-//! let labels = DBSCAN::fit(&x, Distances::euclidian(),
-//!     DBSCANParameters::default().with_eps(3.0)).
+//! let labels = DBSCAN::fit(&x, DBSCANParameters::default().with_eps(3.0)).
 //!     and_then(|dbscan| dbscan.predict(&x));
 //!
 //! println!("{:?}", labels);
@@ -33,9 +32,11 @@ use std::iter::Sum;
 use serde::{Deserialize, Serialize};
 
 use crate::algorithm::neighbour::{KNNAlgorithm, KNNAlgorithmName};
+use crate::api::{Predictor, UnsupervisedEstimator};
 use crate::error::Failed;
 use crate::linalg::{row_iter, Matrix};
-use crate::math::distance::Distance;
+use crate::math::distance::euclidian::Euclidian;
+use crate::math::distance::{Distance, Distances};
 use crate::math::num::RealNumber;
 use crate::tree::decision_tree_classifier::which_max;
 
@@ -50,7 +51,11 @@ pub struct DBSCAN<T: RealNumber, D: Distance<Vec<T>, T>> {
 
 #[derive(Debug, Clone)]
 /// DBSCAN clustering algorithm parameters
-pub struct DBSCANParameters<T: RealNumber> {
+pub struct DBSCANParameters<T: RealNumber, D: Distance<Vec<T>, T>> {
+    /// a function that defines a distance between each pair of point in training data.
+    /// This function should extend [`Distance`](../../math/distance/trait.Distance.html) trait.
+    /// See [`Distances`](../../math/distance/struct.Distances.html) for a list of available functions.
+    pub distance: D,
     /// The number of samples (or total weight) in a neighborhood for a point to be considered as a core point.
     pub min_samples: usize,
     /// The maximum distance between two samples for one to be considered as in the neighborhood of the other.
@@ -59,7 +64,18 @@ pub struct DBSCANParameters<T: RealNumber> {
     pub algorithm: KNNAlgorithmName,
 }
 
-impl<T: RealNumber> DBSCANParameters<T> {
+impl<T: RealNumber, D: Distance<Vec<T>, T>> DBSCANParameters<T, D> {
+    /// a function that defines a distance between each pair of point in training data.
+    /// This function should extend [`Distance`](../../math/distance/trait.Distance.html) trait.
+    /// See [`Distances`](../../math/distance/struct.Distances.html) for a list of available functions.
+    pub fn with_distance<DD: Distance<Vec<T>, T>>(self, distance: DD) -> DBSCANParameters<T, DD> {
+        DBSCANParameters {
+            distance,
+            min_samples: self.min_samples,
+            eps: self.eps,
+            algorithm: self.algorithm,
+        }
+    }
     /// The number of samples (or total weight) in a neighborhood for a point to be considered as a core point.
     pub fn with_min_samples(mut self, min_samples: usize) -> Self {
         self.min_samples = min_samples;
@@ -86,25 +102,41 @@ impl<T: RealNumber, D: Distance<Vec<T>, T>> PartialEq for DBSCAN<T, D> {
     }
 }
 
-impl<T: RealNumber> Default for DBSCANParameters<T> {
+impl<T: RealNumber> Default for DBSCANParameters<T, Euclidian> {
     fn default() -> Self {
         DBSCANParameters {
+            distance: Distances::euclidian(),
             min_samples: 5,
             eps: T::half(),
             algorithm: KNNAlgorithmName::CoverTree,
         }
     }
 }
 
+impl<T: RealNumber + Sum, M: Matrix<T>, D: Distance<Vec<T>, T>>
+    UnsupervisedEstimator<M, DBSCANParameters<T, D>> for DBSCAN<T, D>
+{
+    fn fit(x: &M, parameters: DBSCANParameters<T, D>) -> Result<Self, Failed> {
+        DBSCAN::fit(x, parameters)
+    }
+}
+
+impl<T: RealNumber, M: Matrix<T>, D: Distance<Vec<T>, T>> Predictor<M, M::RowVector>
+    for DBSCAN<T, D>
+{
+    fn predict(&self, x: &M) -> Result<M::RowVector, Failed> {
+        self.predict(x)
+    }
+}
+
 impl<T: RealNumber + Sum, D: Distance<Vec<T>, T>> DBSCAN<T, D> {
     /// Fit algorithm to _NxM_ matrix where _N_ is number of samples and _M_ is number of features.
     /// * `data` - training instances to cluster
     /// * `k` - number of clusters
     /// * `parameters` - cluster parameters
     pub fn fit<M: Matrix<T>>(
         x: &M,
-        distance: D,
-        parameters: DBSCANParameters<T>,
+        parameters: DBSCANParameters<T, D>,
     ) -> Result<DBSCAN<T, D>, Failed> {
         if parameters.min_samples < 1 {
             return Err(Failed::fit(&"Invalid minPts".to_string()));
@@ -121,7 +153,9 @@ impl<T: RealNumber + Sum, D: Distance<Vec<T>, T>> DBSCAN<T, D> {
         let n = x.shape().0;
         let mut y = vec![unassigned; n];
 
-        let algo = parameters.algorithm.fit(row_iter(x).collect(), distance)?;
+        let algo = parameters
+            .algorithm
+            .fit(row_iter(x).collect(), parameters.distance)?;
 
         for (i, e) in row_iter(x).enumerate() {
             if y[i] == unassigned {
@@ -195,7 +229,6 @@ mod tests {
     use super::*;
     use crate::linalg::naive::dense_matrix::DenseMatrix;
     use crate::math::distance::euclidian::Euclidian;
-    use crate::math::distance::Distances;
 
     #[test]
     fn fit_predict_dbscan() {
@@ -215,16 +248,7 @@ mod tests {
 
         let expected_labels = vec![0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, -1.0];
 
-        let dbscan = DBSCAN::fit(
-            &x,
-            Distances::euclidian(),
-            DBSCANParameters {
-                min_samples: 5,
-                eps: 1.0,
-                algorithm: KNNAlgorithmName::CoverTree,
-            },
-        )
-        .unwrap();
+        let dbscan = DBSCAN::fit(&x, DBSCANParameters::default().with_eps(1.0)).unwrap();
 
         let predicted_labels = dbscan.predict(&x).unwrap();
 
@@ -256,7 +280,7 @@ mod tests {
             &[5.2, 2.7, 3.9, 1.4],
         ]);
 
-        let dbscan = DBSCAN::fit(&x, Distances::euclidian(), Default::default()).unwrap();
+        let dbscan = DBSCAN::fit(&x, Default::default()).unwrap();
 
         let deserialized_dbscan: DBSCAN<f64, Euclidian> =
             serde_json::from_str(&serde_json::to_string(&dbscan).unwrap()).unwrap();

src/cluster/dbscan.rs

@@ -43,7 +43,7 @@
 //!            &[5.2, 2.7, 3.9, 1.4],
 //!            ]);
 //!
-//! let kmeans = KMeans::fit(&x, 2, Default::default()).unwrap(); // Fit to data, 2 clusters
+//! let kmeans = KMeans::fit(&x, KMeansParameters::default().with_k(2)).unwrap(); // Fit to data, 2 clusters
 //! let y_hat = kmeans.predict(&x).unwrap(); // use the same points for prediction
 //! ```
 //!
@@ -59,6 +59,7 @@ use std::iter::Sum;
 use serde::{Deserialize, Serialize};
 
 use crate::algorithm::neighbour::bbd_tree::BBDTree;
+use crate::api::{Predictor, UnsupervisedEstimator};
 use crate::error::Failed;
 use crate::linalg::Matrix;
 use crate::math::distance::euclidian::*;
@@ -101,11 +102,18 @@ impl<T: RealNumber> PartialEq for KMeans<T> {
 #[derive(Debug, Clone)]
 /// K-Means clustering algorithm parameters
 pub struct KMeansParameters {
+    /// Number of clusters.
+    pub k: usize,
     /// Maximum number of iterations of the k-means algorithm for a single run.
     pub max_iter: usize,
 }
 
 impl KMeansParameters {
+    /// Number of clusters.
+    pub fn with_k(mut self, k: usize) -> Self {
+        self.k = k;
+        self
+    }
     /// Maximum number of iterations of the k-means algorithm for a single run.
     pub fn with_max_iter(mut self, max_iter: usize) -> Self {
         self.max_iter = max_iter;
@@ -115,24 +123,37 @@ impl KMeansParameters {
 
 impl Default for KMeansParameters {
     fn default() -> Self {
-        KMeansParameters { max_iter: 100 }
+        KMeansParameters {
+            k: 2,
+            max_iter: 100,
+        }
+    }
+}
+
+impl<T: RealNumber + Sum, M: Matrix<T>> UnsupervisedEstimator<M, KMeansParameters> for KMeans<T> {
+    fn fit(x: &M, parameters: KMeansParameters) -> Result<Self, Failed> {
+        KMeans::fit(x, parameters)
+    }
+}
+
+impl<T: RealNumber, M: Matrix<T>> Predictor<M, M::RowVector> for KMeans<T> {
+    fn predict(&self, x: &M) -> Result<M::RowVector, Failed> {
+        self.predict(x)
     }
 }
 
 impl<T: RealNumber + Sum> KMeans<T> {
     /// Fit algorithm to _NxM_ matrix where _N_ is number of samples and _M_ is number of features.
-    /// * `data` - training instances to cluster
-    /// * `k` - number of clusters
+    /// * `data` - training instances to cluster    
     /// * `parameters` - cluster parameters
-    pub fn fit<M: Matrix<T>>(
-        data: &M,
-        k: usize,
-        parameters: KMeansParameters,
-    ) -> Result<KMeans<T>, Failed> {
+    pub fn fit<M: Matrix<T>>(data: &M, parameters: KMeansParameters) -> Result<KMeans<T>, Failed> {
         let bbd = BBDTree::new(data);
 
-        if k < 2 {
-            return Err(Failed::fit(&format!("invalid number of clusters: {}", k)));
+        if parameters.k < 2 {
+            return Err(Failed::fit(&format!(
+                "invalid number of clusters: {}",
+                parameters.k
+            )));
         }
 
         if parameters.max_iter == 0 {
@@ -145,9 +166,9 @@ impl<T: RealNumber + Sum> KMeans<T> {
         let (n, d) = data.shape();
 
         let mut distortion = T::max_value();
-        let mut y = KMeans::kmeans_plus_plus(data, k);
-        let mut size = vec![0; k];
-        let mut centroids = vec![vec![T::zero(); d]; k];
+        let mut y = KMeans::kmeans_plus_plus(data, parameters.k);
+        let mut size = vec![0; parameters.k];
+        let mut centroids = vec![vec![T::zero(); d]; parameters.k];
 
         for i in 0..n {
             size[y[i]] += 1;
@@ -159,16 +180,16 @@ impl<T: RealNumber + Sum> KMeans<T> {
             }
         }
 
-        for i in 0..k {
+        for i in 0..parameters.k {
             for j in 0..d {
                 centroids[i][j] /= T::from(size[i]).unwrap();
             }
         }
 
-        let mut sums = vec![vec![T::zero(); d]; k];
+        let mut sums = vec![vec![T::zero(); d]; parameters.k];
         for _ in 1..=parameters.max_iter {
             let dist = bbd.clustering(&centroids, &mut sums, &mut size, &mut y);
-            for i in 0..k {
+            for i in 0..parameters.k {
                 if size[i] > 0 {
                     for j in 0..d {
                         centroids[i][j] = T::from(sums[i][j]).unwrap() / T::from(size[i]).unwrap();
@@ -184,7 +205,7 @@ impl<T: RealNumber + Sum> KMeans<T> {
         }
 
         Ok(KMeans {
-            k,
+            k: parameters.k,
             y,
             size,
             distortion,
@@ -280,10 +301,10 @@ mod tests {
     fn invalid_k() {
         let x = DenseMatrix::from_2d_array(&[&[1., 2., 3.], &[4., 5., 6.]]);
 
-        assert!(KMeans::fit(&x, 0, Default::default()).is_err());
+        assert!(KMeans::fit(&x, KMeansParameters::default().with_k(0)).is_err());
         assert_eq!(
             "Fit failed: invalid number of clusters: 1",
-            KMeans::fit(&x, 1, Default::default())
+            KMeans::fit(&x, KMeansParameters::default().with_k(1))
                 .unwrap_err()
                 .to_string()
         );
@@ -314,7 +335,7 @@ mod tests {
             &[5.2, 2.7, 3.9, 1.4],
         ]);
 
-        let kmeans = KMeans::fit(&x, 2, Default::default()).unwrap();
+        let kmeans = KMeans::fit(&x, Default::default()).unwrap();
 
         let y = kmeans.predict(&x).unwrap();
 
@@ -348,7 +369,7 @@ mod tests {
             &[5.2, 2.7, 3.9, 1.4],
         ]);
 
-        let kmeans = KMeans::fit(&x, 2, Default::default()).unwrap();
+        let kmeans = KMeans::fit(&x, Default::default()).unwrap();
 
         let deserialized_kmeans: KMeans<f64> =
             serde_json::from_str(&serde_json::to_string(&kmeans).unwrap()).unwrap();