Added Kernel of RBF.

Author: Kalvar

KRKmeans.podspec

@@ -1,6 +1,6 @@
 Pod::Spec.new do |s|
   s.name         = "KRKmeans"
-  s.version      = "2.5.1"
+  s.version      = "2.5.2"
   s.summary      = "KRKmeans is clustering algorithm (クラスタリング分類) that one of Machine Learning methods."
   s.description  = <<-DESC
                    KRKmeans has implemented K-Means the clustering algorithm (クラスタリング分類) and achieved multi-dimensional clustering in this project. KRKmeans could be used in data mining (データマイニング), image compression (画像圧縮) and classification.

KRKmeans.xcodeproj/project.xcworkspace/xcuserdata/Kalvar.xcuserdatad/UserInterfaceState.xcuserstate

@@ -61,7 +61,7 @@ -(void)twoDemensional
     _krKmeans.doneThenSave    = YES;
 
     // Set to use 2 points of Euclidean Distance method that performance is better
-    _krKmeans.distanceFormula = KRKmeansDistanceFormulaByEuclidean; // KRKmeansDistanceFormulaByCosine
+    _krKmeans.distanceFormula = KRKmeansDistanceFormulaEuclidean; // KRKmeansDistanceFormulaByCosine
 
     // It means A sets. ( and the centers will be calculated here. )
     [_krKmeans addSets:@[@[@1, @1], @[@1, @2], @[@2, @2], @[@3, @2], @[@3, @1]]];
@@ -126,7 +126,7 @@ -(void)multiDemensional
     _multiKmeans.doneThenSave = YES;
 
     // Suggests to use Cosine Similarity doing multi-dimensional clustering
-    _multiKmeans.distanceFormula = KRKmeansDistanceFormulaByCosine; // KRKmeansDistanceFormulaByEuclidean
+    _multiKmeans.distanceFormula = KRKmeansDistanceFormulaRBF; // KRKmeansDistanceFormulaCosine; // KRKmeansDistanceFormulaByEuclidean
 
     // A sets
     [_multiKmeans addSets:@[@[@20, @9, @1, @3, @6, @2], @[@52, @32, @18, @7, @0, @1], @[@30, @18, @2, @27, @18, @5]]];
@@ -159,7 +159,7 @@ -(void)multiDemensional
 -(void)directClustering
 {
     KRKmeans *_kmeans       = [KRKmeans sharedKmeans];
-    _kmeans.distanceFormula = KRKmeansDistanceFormulaByEuclidean;
+    _kmeans.distanceFormula = KRKmeansDistanceFormulaEuclidean;
     [_kmeans addPatterns:@[@[@7, @11], @[@18, @6]]];
     [_kmeans directClusterWithCompletion:^(BOOL success, NSArray *clusters, NSArray *centers, NSInteger totalTimes)
     {
@@ -171,7 +171,7 @@ -(void)autoClustering
 {
     KRKmeans *_krKmeans         = [[KRKmeans alloc] init];
     _krKmeans.doneThenSave      = YES;
-    _krKmeans.distanceFormula   = KRKmeansDistanceFormulaByEuclidean; // KRKmeansDistanceFormulaByCosine
+    _krKmeans.distanceFormula   = KRKmeansDistanceFormulaEuclidean; // KRKmeansDistanceFormulaByCosine
     _krKmeans.autoClusterNumber = 3;
     [_krKmeans addPatterns:@[@[@1, @1], @[@1, @2], @[@2, @2], @[@3, @2],
                              @[@3, @1], @[@5, @4], @[@3, @4], @[@2, @5],

KRKmeans/ViewController.m

@@ -1,4 +1,4 @@
-Copyright (c) 2014 - 2015 Kuo-Ming Lin (Kalvar Lin) (http://kalvar-kaki.blogspot.tw/)
+Copyright (c) 2014 - 2016 Kuo-Ming Lin (Kalvar Lin) (http://kalvar-kaki.blogspot.tw/)
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

LICENSE

@@ -37,9 +37,11 @@
 typedef enum KRKmeansDistanceFormula
 {
     // Distance formula by Cosine Similarity
-    KRKmeansDistanceFormulaByCosine    = 0,
+    KRKmeansDistanceFormulaCosine    = 0,
     // Distance formula by Euclidean Distance
-    KRKmeansDistanceFormulaByEuclidean = 1
+    KRKmeansDistanceFormulaEuclidean = 1,
+    // Distance formula by RBF
+    KRKmeansDistanceFormulaRBF       = 2
 }KRKmeansDistanceFormula;
 
 /*
@@ -79,6 +81,8 @@ typedef void(^KRKmeansPerIteration)(NSInteger times, NSArray *clusters, NSArray
 @property (nonatomic, assign) BOOL doneThenSave;
 //Auto clustering numbers (要自動分成幾群)
 @property (nonatomic, assign) NSInteger autoClusterNumber;
+// If we used RBF be the kernel that can setup this considition
+@property (nonatomic, assign) float sigma;
 
 @property (nonatomic, copy) KRKmeansClusteringCompletion clusterCompletion;
 @property (nonatomic, copy) KRKmeansPerIteration perIteration;

ML/KRKmeans.h

@@ -80,7 +80,7 @@ -(NSArray *)_useAverageVectorCalculateCenters:(NSArray *)_clusteredSets
 @implementation KRKmeans (fixDistances)
 
 // Euclidean distance which multi-dimensional formula, 距離越小越近
--(float)_distanceEuclideanX1:(NSArray *)_x1 x2:(NSArray *)_x2
+-(float)_euclideanX1:(NSArray *)_x1 x2:(NSArray *)_x2
 {
     NSInteger _index = 0;
     float _sum       = 0.0f;
@@ -94,7 +94,7 @@ -(float)_distanceEuclideanX1:(NSArray *)_x1 x2:(NSArray *)_x2
 }
 
 // Cosine Similarity method that multi-dimensional, 同歸屬度越大越近
--(float)_distanceCosineSimilarityX1:(NSArray *)_x1 x2:(NSArray *)_x2
+-(float)_cosineSimilarityX1:(NSArray *)_x1 x2:(NSArray *)_x2
 {
     float _sumA  = 0.0f;
     float _sumB  = 0.0f;
@@ -114,17 +114,35 @@ -(float)_distanceCosineSimilarityX1:(NSArray *)_x1 x2:(NSArray *)_x2
     return ( _ab > 0.0f ) ? ( _sumAB / sqrtf( _ab ) ) : 0.0f;
 }
 
+-(double)_rbf:(NSArray *)_x1 x2:(NSArray *)_x2
+{
+    double _sum      = 0.0f;
+    NSInteger _index = 0;
+    for( NSNumber *_value in _x1 )
+    {
+        // Formula : s = s + ( v1[i] - v2[i] )^2
+        double _v  = [_value doubleValue] - [[_x2 objectAtIndex:_index] doubleValue];
+        _sum      += ( _v * _v );
+        ++_index;
+    }
+    // Formula : exp^( -s / ( 2.0f * sigma * sigma ) )
+    return pow(M_E, ((-_sum) / ( 2.0f * self.sigma * self.sigma )));
+}
+
 // 距離概念是越小越近，歸屬度概念是越大越近 ( 或取其差值，使歸屬度同距離越小越近 )
 -(float)_distanceX1:(NSArray *)_x1 x2:(NSArray *)_x2
 {
     float _distance = 0.0f;
     switch (self.distanceFormula)
     {
-        case KRKmeansDistanceFormulaByCosine:
-            _distance = 1.0f - [self _distanceCosineSimilarityX1:_x1 x2:_x2];
+        case KRKmeansDistanceFormulaCosine:
+            _distance = 1.0f - [self _cosineSimilarityX1:_x1 x2:_x2];
             break;
-        case KRKmeansDistanceFormulaByEuclidean:
-            _distance = [self _distanceEuclideanX1:_x1 x2:_x2];
+        case KRKmeansDistanceFormulaEuclidean:
+            _distance = [self _euclideanX1:_x1 x2:_x2];
+            break;
+        case KRKmeansDistanceFormulaRBF:
+            _distance = [self _rbf:_x1 x2:_x2];
             break;
         default:
             break;
@@ -286,8 +304,10 @@ -(instancetype)init
         _lastDistance       = -1.0f;
         _currentIteration   = 0;
 
-        _distanceFormula    = KRKmeansDistanceFormulaByEuclidean;
+        _distanceFormula    = KRKmeansDistanceFormulaEuclidean;
         _trainedSaves       = [KRKmeansSaves sharedInstance];
+        
+        _sigma              = 2.0f;
     }
     return self;
 }

ML/KRKmeans.m

@@ -7,7 +7,7 @@ KRKmeans has implemented K-Means the clustering algorithm (クラスタリング
 
 ```ruby
 platform :ios, '7.0'
-pod "KRKmeans", "~> 2.5.1"
+pod "KRKmeans", "~> 2.5.2"
 ```
 
 ## How to use
@@ -18,6 +18,14 @@ pod "KRKmeans", "~> 2.5.1"
 #import "KRKmeans.h"
 ```
 
+##### Distance Methods
+
+``` objective-c
+KRKmeansDistanceFormulaEuclidean
+KRKmeansDistanceFormulaCosine
+KRKmeansDistanceFormulaRBF
+```
+
 #### One dimensonal clustering
 
 ``` objective-c
@@ -71,7 +79,7 @@ pod "KRKmeans", "~> 2.5.1"
     KRKmeans *_krKmeans = [KRKmeans sharedKmeans];
 
     // Set to Euclidean Distance method that performance is better
-    _krKmeans.dimensional = KRKmeansDistanceFormulaByEuclidean;
+    _krKmeans.dimensional = KRKmeansDistanceFormulaEuclidean;
 
     //It means A sets. ( and the centers will be calculated here. )
     [_krKmeans addSets:@[@[@1, @1], @[@1, @2], @[@2, @2], @[@3, @2], @[@3, @1]]];
@@ -128,7 +136,7 @@ pod "KRKmeans", "~> 2.5.1"
     KRKmeans *_multiKmeans = [[KRKmeans alloc] init];
 
     // Suggests to use Cosine Similarity doing multi-dimensional clustering
-    _multiKmeans.dimensional = KRKmeansDimensionalMultiByCosine;
+    _multiKmeans.dimensional = KRKmeansDistanceFormulaCosine;
 
     // A sets
     [_multiKmeans addSets:@[@[@20, @9, @1, @3, @6, @2], @[@52, @32, @18, @7, @0, @1], @[@30, @18, @2, @27, @18, @5]]];
@@ -167,7 +175,7 @@ If you have trained clusters that you could directly put new patterns into direc
 -(void)directClustering
 {
     KRKmeans *_kmeans   = [KRKmeans sharedKmeans];
-    _kmeans.dimensional = KRKmeansDistanceFormulaByEuclidean;
+    _kmeans.dimensional = KRKmeansDistanceFormulaEuclidean;
     [_kmeans addPatterns:@[@[@7, @11], @[@18, @6]]];
     [_kmeans directClusterWithCompletion:^(BOOL success, NSArray *clusters, NSArray *centers, NSInteger totalTimes)
     {
@@ -185,7 +193,7 @@ Automatic picking the group-centers by your wishes number.
 {
     KRKmeans *_krKmeans         = [[KRKmeans alloc] init];
     _krKmeans.doneThenSave      = YES;
-    _krKmeans.distanceFormula   = KRKmeansDistanceFormulaByEuclidean; // KRKmeansDistanceFormulaByCosine
+    _krKmeans.distanceFormula   = KRKmeansDistanceFormulaEuclidean; // KRKmeansDistanceFormulaCosine
     _krKmeans.autoClusterNumber = 3;
     [_krKmeans addPatterns:@[@[@1, @1], @[@1, @2], @[@2, @2], @[@3, @2],
                              @[@3, @1], @[@5, @4], @[@3, @4], @[@2, @5],
@@ -220,7 +228,7 @@ Automatic picking the group-centers by your wishes number.
 
 ## Version
 
-V2.5.1
+V2.5.2
 
 ## License