More KNN experiments

2019-09-05 18:34:50 -07:00
parent a9ec6dfcd0
commit 9c5f6eb307
5 changed files with 118 additions and 63 deletions
@@ -1,31 +1,84 @@
 use super::Classifier;
-use crate::math::distance::Distance;
-use crate::math::distance::euclidian::EuclidianDistance;
 use crate::algorithm::sort::heap_select::HeapSelect;
+use crate::common::AnyNumber;
 use ndarray::prelude::*;
-use num_traits::Signed;
-use num_traits::{Float, Num};
-use std::marker::PhantomData;
+use ndarray::{ArrayBase, Data, Ix1, Ix2};
+use num_traits::{Float};
 use std::cmp::{Ordering, PartialOrd};
-use std::fmt::Debug;
+use ndarray::arr1;

-pub struct KNNClassifier<E> {
-    y: Option<Array1<E>>
+pub struct KNNClassifier<X, Y, F> 
+where
+    X: AnyNumber,
+    Y: AnyNumber,    
+    F: Fn(&Array1<X>, &Array1<X>) -> f64
+{     
+    y: Vec<Y>,  
+    distance: F,
+    k: usize,
+    knn_algorithm: Box<KNNAlgorithm<Array1<X>, F>>    
 }

-pub trait KNNAlgorithm<T: Clone + Debug>{
-    fn find(&self, from: &T, k: usize) -> Vec<&T>;
-}
-
-pub struct SimpleKNNAlgorithm<T, D: Distance<T>>
+impl<X, Y, F> KNNClassifier<X, Y, F> 
+where
+    X: AnyNumber,
+    Y: AnyNumber,    
+    F: Fn(&Array1<X>, &Array1<X>) -> f64
 {
-    data: Vec<T>,
-    distance: D  
+
+    pub fn fit<SX: Data<Elem = X>, SY: Data<Elem = Y>>(x: &ArrayBase<SX, Ix2>, y: &ArrayBase<SY, Ix1>, k: usize, distance: F) -> KNNClassifier<X, Y, F> {
+
+        assert!(ArrayBase::shape(x)[0] == ArrayBase::shape(y)[0], format!("Size of x should equal size of y; |x|=[{}], |y|=[{}]", ArrayBase::shape(x)[0], ArrayBase::shape(y)[0]));
+
+        assert!(k > 1, format!("k should be > 1, k=[{}]", k));        
+                  
+        let v: Vec<Array1<X>> = x.outer_iter().map(|x| x.to_owned()).collect(); 
+
+        let knn = Box::new(SimpleKNNAlgorithm{
+            data: v
+        });       
+
+        KNNClassifier{y: y.to_owned().to_vec(), k: k, distance: distance, knn_algorithm: knn}
+    }
 }

-impl<T: Clone + Debug, D: Distance<T>> KNNAlgorithm<T> for SimpleKNNAlgorithm<T, D>
+impl<X, Y, SX, F> Classifier<X, Y, SX> for KNNClassifier<X, Y, F> 
+where    
+    X: AnyNumber,
+    Y: AnyNumber,
+    SX: Data<Elem = X>,    
+    F: Fn(&Array1<X>, &Array1<X>) -> f64
+ {    
+
+    fn predict(&self, x: &ArrayBase<SX, Ix2>) -> Array1<Y> {             
+        let mut result = Vec::new();
+        for x in x.outer_iter() {
+            let idxs = self.knn_algorithm.find(&x.to_owned(), self.k, &self.distance);
+            let mut sum: Y = Y::zero();
+            let mut count = 0;
+            for i in idxs {
+                sum = sum + self.y[i].to_owned();
+                count += 1;
+            }
+            result.push(sum / Y::from_u64(count).unwrap());
+        }          
+        arr1(&result)
+    }
+
+}
+
+pub trait KNNAlgorithm<T: Clone, F: Fn(&T, &T) -> f64>{
+    fn find(&self, from: &T, k: usize, d: &F) -> Vec<usize>;
+}
+
+pub struct SimpleKNNAlgorithm<T>
 {
-    fn find(&self, from: &T, k: usize) -> Vec<&T> {
+    data: Vec<T> 
+}
+
+impl<T: Clone, F: Fn(&T, &T) -> f64> KNNAlgorithm<T, F> for SimpleKNNAlgorithm<T>
+{
+    fn find(&self, from: &T, k: usize, d: &F) -> Vec<usize> {
        if k < 1 || k > self.data.len() {
            panic!("k should be >= 1 and <= length(data)");
        }        
@@ -41,7 +94,7 @@ impl<T: Clone + Debug, D: Distance<T>> KNNAlgorithm<T> for SimpleKNNAlgorithm<T,

        for i in 0..self.data.len() {

-            let d = D::distance(&from, &self.data[i]);
+            let d = d(&from, &self.data[i]);
            let datum = heap.peek_mut();            
            if d < datum.distance {
                datum.distance = d;
@@ -52,7 +105,7 @@ impl<T: Clone + Debug, D: Distance<T>> KNNAlgorithm<T> for SimpleKNNAlgorithm<T,

        heap.sort(); 

-        heap.get().into_iter().flat_map(|x| x.index).map(|i| &self.data[i]).collect()
+        heap.get().into_iter().flat_map(|x| x.index).collect()
    }
 }

@@ -76,24 +129,11 @@ impl PartialEq for KNNPoint {

 impl Eq for KNNPoint {}

-impl<A1, A2>  Classifier<A1, A2> for KNNClassifier<A2> 
-where    
-    A2: Signed + Clone,    
- {
-    fn fit(&mut self, x: &Array2<A1>, y: &Array1<A2>){ 
-          self.y = Some(Array1::<A2>::zeros(ArrayBase::len(y)));           
-    }
-
-    fn predict(&self, x: &Array2<A1>) -> Array1<A2>{             
-        let array = Array1::<A2>::zeros(ArrayBase::len(self.y.as_ref().unwrap()));
-        array
-    }
-
-}
-
 #[cfg(test)]
 mod tests {    
-    use super::*;   
+    use super::*;  
+    use crate::math::distance::Distance;
+    use crate::math::distance::euclidian::EuclidianDistance; 

    struct SimpleDistance{}

@@ -104,23 +144,28 @@ mod tests {
    }     

    #[test]
-    fn knn_fit_predict() {        
-        let mut knn = KNNClassifier{y: None};
-        let x = arr2(&[[1, 2, 3],[4, 5, 6]]);        
-        let y = arr1(&[1, 2]);
-        knn.fit(&x, &y);
+    fn knn_fit_predict() {                
+        let x = arr2(&[[1, 2], [3, 4], [5, 6], [7, 8], [9, 10]]);        
+        let y = arr1(&[1, 2, 3, 4, 5]);
+        let knn = KNNClassifier::fit(&x, &y, 3, EuclidianDistance::distance);
        let r = knn.predict(&x);
-        assert_eq!(2, ArrayBase::len(&r));
+        assert_eq!(5, ArrayBase::len(&r));
+        assert_eq!(arr1(&[2, 2, 3, 4, 4]), r);
    }

    #[test]
    fn knn_find() {        
-        let sKnn = SimpleKNNAlgorithm{
-            data: vec!(1, 2, 3, 4, 5, 6, 7, 8, 9, 10),
-            distance: SimpleDistance{}    
-        };        
+        let simple_knn = SimpleKNNAlgorithm{
+            data: vec!(1, 2, 3, 4, 5, 6, 7, 8, 9, 10)
+        };            

-        assert_eq!(vec!(&2, &3, &1), sKnn.find(&2, 3));
+        assert_eq!(vec!(1, 2, 0), simple_knn.find(&2, 3, &SimpleDistance::distance));
+
+        let knn2 = SimpleKNNAlgorithm{
+            data: vec!(arr1(&[1, 1]), arr1(&[2, 2]), arr1(&[3, 3]), arr1(&[4, 4]), arr1(&[5, 5]))
+        }; 
+
+        assert_eq!(vec!(2, 3, 1), knn2.find(&arr1(&[3, 3]), 3, &EuclidianDistance::distance));
    }

    #[test]
@@ -1,14 +1,15 @@
-use ndarray::prelude::*;
-use ndarray::{arr1, arr2};
-use ndarray::FixedInitializer;
+use crate::common::AnyNumber;
+use ndarray::{Array1, ArrayBase, Data, Ix2};

 pub mod knn;

-pub trait Classifier<E1, E2>
-{
+pub trait Classifier<X, Y, SX>
+where 
+    X: AnyNumber,
+    Y: AnyNumber,
+    SX: Data<Elem = X>    
+{    

-    fn fit(&mut self, x: &Array2<E1>, y: &Array1<E2>);
-
-    fn predict(&self, x: &Array2<E1>) -> Array1<E2>;
+    fn predict(&self, x: &ArrayBase<SX, Ix2>) -> Array1<Y>;

 }
@@ -0,0 +1,7 @@
+use num_traits::{Num, ToPrimitive, FromPrimitive};
+use ndarray::{ScalarOperand};
+
+pub trait AnyNumber: Num + ScalarOperand + ToPrimitive + FromPrimitive{}
+
+
+impl<T> AnyNumber for T where T: Num + ScalarOperand + ToPrimitive + FromPrimitive {}
@@ -1,4 +1,5 @@
 pub mod classification;
 pub mod math;
 pub mod error;
-pub mod algorithm;
+pub mod algorithm;
+pub mod common;
@@ -1,13 +1,12 @@
 use super::Distance;
 use ndarray::{ArrayBase, Data, Dimension};
-use num_traits::{Num, ToPrimitive};
-use ndarray::{ScalarOperand};
+use crate::common::AnyNumber;

 pub struct EuclidianDistance{}

 impl<A, S, D> Distance<ArrayBase<S, D>> for EuclidianDistance
 where
-    A: Num + ScalarOperand + ToPrimitive,      
+    A: AnyNumber,      
    S: Data<Elem = A>,
    D: Dimension
 {
@@ -25,17 +24,19 @@ where
 #[cfg(test)]
 mod tests {
    use super::*;
-    use ndarray::{arr1, Array};    
+    use ndarray::arr1;    

    #[test]
    fn measure_simple_euclidian_distance() {
        let a = arr1(&[1, 2, 3]);
-        let b = arr1(&[4, 5, 6]);        
+        let b = arr1(&[4, 5, 6]);             

-        let d = EuclidianDistance::distance(&a, &b);
+        let d_arr = EuclidianDistance::distance(&a, &b);
+        let d_view = EuclidianDistance::distance(&a.view(), &b.view());

-        assert!((d - 5.19615242).abs() < 1e-8);
-    }
+        assert!((d_arr - 5.19615242).abs() < 1e-8);
+        assert!((d_view - 5.19615242).abs() < 1e-8);
+    }    

    #[test]
    fn measure_simple_euclidian_distance_static() {