From 039ff4a04718d459acb06b29ca74c73c0ab3ba3f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E2=8E=9B=E2=8E=9D=20L=E2=88=86RBI=20=E2=8E=A0=E2=8E=9E?=
 <48409226+L4RBI@users.noreply.github.com>
Date: Thu, 1 Oct 2020 02:08:07 +0100
Subject: [PATCH] added a classification algorithms folder

---
 classification/fcm.py   | 57 +++++++++++++++++++++++++++++++++++++++++
 classification/tools.py | 20 +++++++++++++++
 2 files changed, 77 insertions(+)
 create mode 100644 classification/fcm.py
 create mode 100644 classification/tools.py

diff --git a/classification/fcm.py b/classification/fcm.py
new file mode 100644
index 0000000..79a3ee9
--- /dev/null
+++ b/classification/fcm.py
@@ -0,0 +1,57 @@
+from tools import *
+
+# https://en.wikipedia.org/wiki/Fuzzy_clustering
+
+
+class FuzzyCMeans:
+    def __init__(self, n_clusters, initial_centers, data, max_iter=250, m=2, error=1e-5):
+        assert m > 1
+        #assert initial_centers.shape[0] == n_clusters
+        self.U = None
+        self.centers = initial_centers
+        self.max_iter = max_iter
+        self.m = m
+        self.error = error
+        self.data = data
+
+    def membership(self, data, centers):
+        U_temp = cdist(data, centers, 'euclidean')
+        U_temp = numpy.power(U_temp, 2/(self.m - 1))
+        denominator_ = U_temp.reshape(
+            (data.shape[0], 1, -1)).repeat(U_temp.shape[-1], axis=1)
+        denominator_ = U_temp[:, :, numpy.newaxis] / denominator_
+        return 1 / denominator_.sum(2)
+
+    def Centers(self, data, U):
+        um = U ** self.m
+        return (data.T @ um / numpy.sum(um, axis=0)).T
+
+    def newImage(self, U, centers, im):
+        best = numpy.argmax(self.U, axis=-1)
+        # print(best)
+        # numpy.round()
+        image = im.astype(int)
+        for i in range(256):
+            image = numpy.where(image == float(i), centers[best[i]][0], image)
+        return image
+
+    def compute(self):
+        self.U = self.membership(self.data, self.centers)
+
+        past_U = numpy.copy(self.U)
+        begin_time = datetime.datetime.now()
+        for i in range(self.max_iter):
+
+            self.centers = self.Centers(self.data, self.U)
+            self.U = self.membership(self.data, self.centers)
+
+            if norm(self.U - past_U) < self.error:
+                break
+            past_U = numpy.copy(self.U)
+        x = datetime.datetime.now() - begin_time
+        return self.centers, self.U, x
+
+# that's how you run it, data being your data, and the other parameters being the basic FCM parameters such as numbe rof cluseters, degree of fuzziness and so on
+# f = FuzzyCMeans(n_clusters=C, initial_centers=Initial_centers,
+#                 data=data m=2, max_iter=1000, error=1e-5)
+# centers, U, time = f.compute()
diff --git a/classification/tools.py b/classification/tools.py
new file mode 100644
index 0000000..682268a
--- /dev/null
+++ b/classification/tools.py
@@ -0,0 +1,20 @@
+from matplotlib.image import imread
+import matplotlib.pyplot as plt
+from math import sqrt
+import math
+import random
+import numpy
+import operator
+from scipy.spatial.distance import cdist
+from scipy.linalg import norm
+import datetime
+
+
+def Histogram(path):
+    image = imread(path)
+    if len(image.shape) != 2:
+        def gray(rgb): return numpy.dot(rgb[..., :3], [0.2989, 0.5870, 0.1140])
+        gray = gray(image)
+        image = gray
+    hist, bins = numpy.histogram(image.ravel(), 256, [0, 256])
+    return adapt(hist)