Log Regression Classifier Model and Training Algorithm

Classifier.py

import numpy as np

class classifier():
    
    def __init__(self, lr=0.001, num_dims=6):
        self.lr = lr
        self.params = {}
        
        self.W = np.random.rand(num_dims, 1)
        self.b = np.ones((1,1))
        self.dW = np.zeros_like(self.W.shape)
        self.db = np.zeros_like(self.b.shape)
        
        self.metrics = {'error': []}
        
    def fit(self, x, Y, predict=False):
        
        m = x.shape[0]
        W = self.W
        b = self.b
        
        z = np.dot(x, W) + b
        a = 1 / (1 + np.exp(-z))
        
        if predict: return a
        
        C = (-np.dot(np.log(a).T, Y) - np.dot(np.log(1 - a).T, (1 - Y))) / m
        self.metrics['error'] += [C]
        
        dz = a - Y
        
        self.dW = np.dot(x.T, dz)
        self.db = np.sum(dz) / m
        
        self.W -= self.lr * self.dW
        self.b -= self.lr * self.db
        
model = classifier()

m = x.shape[0]
epoch = range(20)

hist = {'epoch': [], 'error': []}

for e in epoch:
    for i in range(10, m, 10):
        x_batch = x_train[i-10:i]
        y_batch = y_train[i-10:i]
        model.fit(x_batch, y_batch)
        
    hist['epoch'] += [e]
    hist['error'] += [np.mean(model.metrics['error'])]
    model.metrics['error'] = []