Fix entroy and ig

fimdlp/CPPFImdlp.cpp

@@ -1,6 +1,7 @@
 #include "CPPFImdlp.h"
 #include <numeric>
 #include <iostream>
+#include <stdio.h>
 #include "Metrics.h"
 namespace CPPFImdlp
 {
@@ -20,7 +21,8 @@ namespace CPPFImdlp
         std::vector<float> cutPts;
         std::vector<int> cutIdx;
         float xPrev, cutPoint;
-        int yPrev, idxPrev;
+        int yPrev;
+        size_t idxPrev;
         std::vector<size_t> indices = sortIndices(X);
         xPrev = X.at(indices[0]);
         yPrev = y.at(indices[0]);
@@ -34,7 +36,7 @@ namespace CPPFImdlp
             //  Definition 2 Cut points are always on boundaries
             if (y.at(*index) != yPrev && xPrev < X.at(*index))
             {
-                cutPoint = round((X.at(*index) + xPrev) / 2 * divider) / divider;
+                cutPoint = round(divider * (X.at(*index) + xPrev) / 2) / divider;
                 if (debug)
                 {
                     std::cout << "Cut point: " << (xPrev + X.at(*index)) / 2 << " //";
@@ -57,6 +59,13 @@ namespace CPPFImdlp
             std::cout << *cutPoint << " -> " << Metrics::informationGain(y, indices, 0, indices.size(), *cutPoint, nc) << std::endl;
             //  << Metrics::informationGain(y, 0, y.size(), *cutPoint, Metrics::numClasses(y, 0, y.size())) << std::endl;
         }
+        std::cout << "+++++++++++++++++++++++" << std::endl;
+        for (size_t i = 0; i < y.size(); i++)
+        {
+            printf("(%3.1f, %d)\n", X[indices.at(i)], y[indices.at(i)]);
+        }
+        std::cout << "+++++++++++++++++++++++" << std::endl;
+
         return cutPts;
     }
     // Argsort from https://stackoverflow.com/questions/1577475/c-sorting-and-keeping-track-of-indexes

fimdlp/Metrics.cpp

@@ -4,7 +4,7 @@ namespace CPPFImdlp
     Metrics::Metrics()
     {
     }
-    int Metrics::numClasses(std::vector<int> &y, std::vector<size_t> indices, int start, int end)
+    int Metrics::numClasses(std::vector<int> &y, std::vector<size_t> indices, size_t start, size_t end)
     {
         int nClasses = 1;
         int yAnt = y.at(start);
@@ -18,7 +18,7 @@ namespace CPPFImdlp
         }
         return nClasses;
     }
-    float Metrics::entropy(std::vector<int> &y, std::vector<size_t> &indices, int start, int end, int nClasses)
+    float Metrics::entropy(std::vector<int> &y, std::vector<size_t> &indices, size_t start, size_t end, int nClasses)
     {
         float entropy = 0;
         int nElements = 0;
@@ -38,7 +38,7 @@ namespace CPPFImdlp
         }
         return entropy;
     }
-    float Metrics::informationGain(std::vector<int> &y, std::vector<size_t> &indices, int start, int end, int cutPoint, int nClasses)
+    float Metrics::informationGain(std::vector<int> &y, std::vector<size_t> &indices, size_t start, size_t end, size_t cutPoint, int nClasses)
     {
         float iGain = 0.0;
         float entropy, entropyLeft, entropyRight;

fimdlp/Metrics.h

@@ -9,9 +9,9 @@ namespace CPPFImdlp
     {
     public:
         Metrics();
-        static int numClasses(std::vector<int> &, std::vector<size_t>, int, int);
-        static float entropy(std::vector<int> &, std::vector<size_t> &, int, int, int);
-        static float informationGain(std::vector<int> &y, std::vector<size_t> &indices, int start, int end, int cutPoint, int nClasses);
+        static int numClasses(std::vector<int> &, std::vector<size_t>, size_t, size_t);
+        static float entropy(std::vector<int> &, std::vector<size_t> &, size_t, size_t, int);
+        static float informationGain(std::vector<int> &, std::vector<size_t> &, size_t, size_t, size_t, int);
     };
 }
 #endif

fimdlp/mdlp.py

@@ -95,13 +95,21 @@ class FImdlp(TransformerMixin, BaseEstimator):
         print("Cut points for each feature in Iris dataset:")
         yz = self.y_.copy()
         xz = X[:, 0].copy()
-        print("Xz: ", xz)
-        print("Yz: ", yz)
-        print("Solución:")
-        print("Xz*: ", np.sort(X[:, 0]))
-        print("yz*: ", yz[np.argsort(X[:, 0])])
+        xz = xz[np.argsort(X[:, 0])]
+        yz = yz[np.argsort(X[:, 0])]
+        cuts = []
+        for i in range(1, len(yz)):
+            if yz[i] != yz[i - 1] and xz[i - 1] < xz[i]:
+                print(f"Cut point: ({xz[i-1]}, {xz[i]}) ({yz[i-1]}, {yz[i]})")
+                cuts.append((xz[i] + xz[i - 1]) / 2)
         for i in range(0, 1):  # self.n_features_):
             datax = np.sort(X[:, i])
             Xcutpoints = self.discretizer_.cut_points(datax, self.y_)
             print(f"{self.features_[i]:20s}: {Xcutpoints}")
+        print("Solución cut_points: ", cuts)
+        print(xz)
+        print("***********")
+        for i in range(0, len(yz)):
+            print(f"({xz[i]}, {yz[i]})")
+        print("***********")
         return X

sample.py

@@ -7,7 +7,9 @@ X = data.data
 y = data.target
 features = data.feature_names
 test = FImdlp()
-# Xcutpoints = test.fit(X, y, features=features).transform(X)
+Xcutpoints = test.fit(X, y, features=features).transform(X)
 clf = CFImdlp(debug=True)
 print("Cut points for feature 0 in Iris dataset:")
 print(clf.cut_points(X[:, 0], y))
+print("Xcut")
+print(Xcutpoints)

setup.py

@@ -17,7 +17,7 @@ setup(
             ],
             language="c++",
             include_dirs=["fimdlp"],
-            extra_compile_args=["-std=c++20"],
+            extra_compile_args=["-std=c++2a"],
         ),
     ]
 )