Add fit_with_centroids

2023-03-21 17:37:58 +09:00
parent 0e1bf6ce7f
commit 8765bd2173
1 changed files with 177 additions and 1 deletions
@@ -62,7 +62,7 @@ use serde::{Deserialize, Serialize};
 use crate::algorithm::neighbour::bbd_tree::BBDTree;
 use crate::api::{Predictor, UnsupervisedEstimator};
 use crate::error::Failed;
-use crate::linalg::basic::arrays::{Array1, Array2};
+use crate::linalg::basic::arrays::{Array1, Array2, Array};
 use crate::metrics::distance::euclidian::*;
 use crate::numbers::basenum::Number;
 use crate::rand_custom::get_rng_impl;
@@ -322,6 +322,109 @@ impl<TX: Number, TY: Number, X: Array2<TX>, Y: Array1<TY>> KMeans<TX, TY, X, Y>
        })
    }

+    /// Fit algorithm to _NxM_ matrix where _N_ is number of samples and _M_ is number of features.
+    /// * `data` - training instances to cluster    
+    /// * `parameters` - cluster parameters
+    /// * `centroids` - starting centroids
+    pub fn fit_with_centroids(
+        data: &X,
+        parameters: KMeansParameters,
+        centroids: Vec<Vec<f64>>,
+    ) -> Result<KMeans<TX, TY, X, Y>, Failed> {
+
+        // TODO: reuse existing methods in `crate::metrics`
+        fn euclidean_distance(point1: &Vec<f64>, point2: &Vec<f64>) -> f64 {
+            let mut dist = 0.0;
+            for i in 0..point1.len() {
+                dist += (point1[i] - point2[i]).powi(2);
+            }
+            dist.sqrt()
+        }
+
+        fn closest_centroid(point: &Vec<f64>, centroids: &Vec<Vec<f64>>) -> usize {
+            let mut closest_idx = 0;
+            let mut closest_dist = std::f64::MAX;
+            for (i, centroid) in centroids.iter().enumerate() {
+                let dist = euclidean_distance(point, centroid);
+                if dist < closest_dist {
+                    closest_dist = dist;
+                    closest_idx = i;
+                }
+            }
+            closest_idx
+        }
+
+        let bbd = BBDTree::new(data);
+    
+        if centroids.len() != parameters.k {
+            return Err(Failed::fit(&format!(
+                "number of centroids ({}) must be equal to k ({})",
+                centroids.len(),
+                parameters.k
+            )));
+        }
+    
+        let mut y = vec![0; data.shape().0];
+        for i in 0..data.shape().0 {
+            y[i] = closest_centroid(
+                &Vec::from_iterator(data.get_row(i).iterator(0).map(|e| e.to_f64().unwrap()),
+                data.shape().1), &centroids
+            );
+        }
+    
+        let mut size = vec![0; parameters.k];
+        let mut new_centroids = vec![vec![0f64; data.shape().1]; parameters.k];
+    
+        for i in 0..data.shape().0 {
+            size[y[i]] += 1;
+        }
+    
+        for i in 0..data.shape().0 {
+            for j in 0..data.shape().1 {
+                new_centroids[y[i]][j] += data.get((i, j)).to_f64().unwrap();
+            }
+        }
+    
+        for i in 0..parameters.k {
+            for j in 0..data.shape().1 {
+                new_centroids[i][j] /= size[i] as f64;
+            }
+        }
+    
+        let mut sums = vec![vec![0f64; data.shape().1]; parameters.k];
+        let mut distortion = std::f64::MAX;
+    
+        for _ in 1..=parameters.max_iter {
+            let dist = bbd.clustering(&new_centroids, &mut sums, &mut size, &mut y);
+            for i in 0..parameters.k {
+                if size[i] > 0 {
+                    for j in 0..data.shape().1 {
+                        new_centroids[i][j] = sums[i][j] / size[i] as f64;
+                    }
+                }
+            }
+    
+            if distortion <= dist {
+                break;
+            } else {
+                distortion = dist;
+            }
+        }
+    
+        Ok(KMeans {
+            k: parameters.k,
+            _y: y,
+            size,
+            _distortion: distortion,
+            centroids: new_centroids,
+            _phantom_tx: PhantomData,
+            _phantom_ty: PhantomData,
+            _phantom_x: PhantomData,
+            _phantom_y: PhantomData,
+        })
+    }
+
+
    /// Predict clusters for `x`
    /// * `x` - matrix with new data to transform of size _KxM_ , where _K_ is number of new samples and _M_ is number of features.
    pub fn predict(&self, x: &X) -> Result<Y, Failed> {
@@ -417,6 +520,7 @@ impl<TX: Number, TY: Number, X: Array2<TX>, Y: Array1<TY>> KMeans<TX, TY, X, Y>
 mod tests {
    use super::*;
    use crate::linalg::basic::matrix::DenseMatrix;
+    use crate::algorithm::neighbour::fastpair;

    #[cfg_attr(
        all(target_arch = "wasm32", not(target_os = "wasi")),
@@ -503,6 +607,78 @@ mod tests {
        }
    }

+    #[cfg_attr(
+        all(target_arch = "wasm32", not(target_os = "wasi")),
+        wasm_bindgen_test::wasm_bindgen_test
+    )]
+    #[test]
+    fn fit_with_centroids_predict() {
+        let x = DenseMatrix::from_2d_array(&[
+            &[5.1, 3.5, 1.4, 0.2],
+            &[4.9, 3.0, 1.4, 0.2],
+            &[4.7, 3.2, 1.3, 0.2],
+            &[4.6, 3.1, 1.5, 0.2],
+            &[5.0, 3.6, 1.4, 0.2],
+            &[5.4, 3.9, 1.7, 0.4],
+            &[4.6, 3.4, 1.4, 0.3],
+            &[5.0, 3.4, 1.5, 0.2],
+            &[4.4, 2.9, 1.4, 0.2],
+            &[4.9, 3.1, 1.5, 0.1],
+            &[7.0, 3.2, 4.7, 1.4],
+            &[6.4, 3.2, 4.5, 1.5],
+            &[6.9, 3.1, 4.9, 1.5],
+            &[5.5, 2.3, 4.0, 1.3],
+            &[6.5, 2.8, 4.6, 1.5],
+            &[5.7, 2.8, 4.5, 1.3],
+            &[6.3, 3.3, 4.7, 1.6],
+            &[4.9, 2.4, 3.3, 1.0],
+            &[6.6, 2.9, 4.6, 1.3],
+            &[5.2, 2.7, 3.9, 1.4],
+        ]);
+
+        let parameters = KMeansParameters {
+            k: 3,
+            max_iter: 50,
+            ..Default::default()
+        };
+
+        // compute pairs
+        let fastpair = fastpair::FastPair::new(&x).unwrap();
+
+        // compute centroids for N closest pairs
+        let mut n: isize = 2;
+        let mut centroids = vec![vec![0f64; x.shape().1]; n as usize + 1];
+        for p in fastpair.ordered_pairs() {
+            if n == -1 {
+                break
+            }
+
+            centroids[n as usize] = {
+                let mut result: Vec<f64> = Vec::with_capacity(x.shape().1);
+                for val1 in x.get_row(p.node).iterator(0) {
+                    for val2 in x.get_row(p.neighbour.unwrap()).iterator(0) {
+                        let sum = val1 + val2;
+                        let avg = sum * 0.5f64;
+                        result.push(avg);
+                    }
+                }
+                result
+            };
+
+            n -= 1;
+        }
+
+
+        let kmeans = KMeans::fit_with_centroids(
+            &x, parameters, centroids).unwrap();
+
+        let y: Vec<usize> = kmeans.predict(&x).unwrap();
+
+        for (i, _y_i) in y.iter().enumerate() {
+            assert_eq!({ y[i] }, kmeans._y[i]);
+        }
+    }
+
    #[cfg_attr(
        all(target_arch = "wasm32", not(target_os = "wasi")),
        wasm_bindgen_test::wasm_bindgen_test