Initial commit

ML/algorithms/logisticregression/logistic_regression.py

@@ -0,0 +1,70 @@
+"""
+From scratch implementation of Logistic Regression
+
+Programmed by Aladdin Persson <aladdin.persson at hotmail dot com>
+*    2020-05-24 Initial coding
+
+"""
+
+import numpy as np
+from sklearn.datasets import make_blobs
+
+
+class LogisticRegression:
+    def __init__(self, X, learning_rate=0.1, num_iters=10000):
+        self.lr = learning_rate
+        self.num_iters = num_iters
+
+        # m for #training_examples, n for #features
+        self.m, self.n = X.shape
+
+    def train(self, X, y):
+        # init weights
+        self.weights = np.zeros((self.n, 1))
+        self.bias = 0
+
+        for it in range(self.num_iters + 1):
+            # calculate hypothesis
+            y_predict = self.sigmoid(np.dot(X, self.weights) + self.bias)
+
+            # calculate cost
+            cost = (
+                -1
+                / self.m
+                * np.sum(y * np.log(y_predict) + (1 - y) * np.log(1 - y_predict))
+            )
+
+            # back prop / gradient calculations
+            dw = 1 / self.m * np.dot(X.T, (y_predict - y))
+            db = 1 / self.m * np.sum(y_predict - y)
+
+            # gradient descent update step
+            self.weights -= self.lr * dw
+            self.bias -= self.lr * db
+
+            # print cost sometimes
+            if it % 1000 == 0:
+                print(f"Cost after iteration {it}: {cost}")
+
+        return self.weights, self.bias
+
+    def predict(self, X):
+        y_predict = self.sigmoid(np.dot(X, self.weights) + self.bias)
+        y_predict_labels = y_predict > 0.5
+
+        return y_predict_labels
+
+    def sigmoid(self, z):
+        return 1 / (1 + np.exp(-z))
+
+
+if __name__ == "__main__":
+    np.random.seed(1)
+    X, y = make_blobs(n_samples=1000, centers=2)
+    y = y[:, np.newaxis]
+
+    logreg = LogisticRegression(X)
+    w, b = logreg.train(X, y)
+    y_predict = logreg.predict(X)
+
+    print(f"Accuracy: {np.sum(y==y_predict)/X.shape[0]}")