fork download
copy 
  1. class MyLogisticRegression:
  2.  
  3.     def __init__(self, learning_rate=0.01, max_iter=1000):
  4.         self.learning_rate = learning_rate
  5.         self.max_iter = max_iter
  6.         self.weights = None  
  7.         self.bias = 0 
  8.         self.all_weights = []
  9.         self.all_biases = []
  10.         self.unique_classes = 0
  11.  
  12.     def initialize_weights(self, n_features):
  13.         self.weights = np.zeros(n_features)
  14.         self.bias = 0  
  15.  
  16.     def sigmoid(self, z):
  17.         z = np.clip(z, -500, 500)
  18.         return 1 / (1 + np.exp(-z))
  19.  
  20.     def get_gradient(self, x, y):
  21.         m = x.shape[0]
  22.         predictions = x @ self.weights + self.bias 
  23.         errors = self.sigmoid(predictions) - y
  24.         weight_gradient = x.T @ errors 
  25.         bias_gradient = np.sum(errors)
  26.         return weight_gradient / m, bias_gradient / m
  27.  
  28.  
  29.     def binary_fit(self, x_train, y_train):
  30.         rows = x_train.shape[0]
  31.         cols = x_train.shape[1]
  32.  
  33.         if self.weights is None: 
  34.             self.initialize_weights(cols)
  35.  
  36.         for i in range(self.max_iter):
  37.             gradient_weights, gradient_bias = self.get_gradient(x_train, y_train)
  38.             self.weights = self.weights - self.learning_rate * gradient_weights
  39.             self.bias = self.bias - self.learning_rate * gradient_bias
  40.         return self.weights, self.bias
  41.  
  42.     def fit(self, x_train, y_train):
  43.  
  44.         self.unique_classes = len(np.unique(y_train))
  45.  
  46.         for i in range(self.unique_classes):
  47.             new_y = (y_train == i).astype(int)
  48.             self.weights = None
  49.             weights, bias = self.binary_fit(x_train, new_y)
  50.             self.all_weights.append(weights)
  51.             self.all_biases.append(bias)
  52.  
  53.     def predict_single_class(self, x_test, weights, bias):  
  54.         prediction =  np.dot(x_test, weights) + bias
  55.         binary_predictions = self.sigmoid(prediction)
  56.         return binary_predictions  
  57.  
  58.     def predict(self, x_test):
  59.         predict_prob = np.zeros((x_test.shape[0], self.unique_classes))
  60.         for i in range(self.unique_classes):
  61.             predict_prob[:, i] = self.predict_single_class(x_test, self.all_weights[i], self.all_biases[i])
  62.         return np.argmax(predict_prob, axis=1)
Success #stdin #stdout 0.1s 14152KB
comments (?)
stdin
copy 
Standard input is empty
stdout
copy 
Standard output is empty
https://ideone.com/CupBHx
language:
Python 3 (python  3.12)
created:
2 days ago
visibility:
public

Share or Embed source code

Discover > Sphere Engine API

The brand new service which powers Ideone!

Discover > IDE Widget

Widget for compiling and running the source code in a web browser!