python复现感知机的二分类算法

本博客复现感知机的二分类算法

import pandas as pd
import numpy as np


class Perceptron:

    def __init__(self, w=None, b=0):
        self.w = w
        self.b = b

    def fit(self, x_train: pd.DataFrame, y_train: pd.Series, iter=20, learn_rate=0.001):
        """
        :param x_train: 训练集特征
        :param y_train: 训练集分类
        :param iter: 迭代次数
        :param learn_rate: 学习率
        :return:
        """
        assert x_train.shape[0] == y_train.shape[0]
        w = np.zeros((1, x_train.shape[1]))
        b = 0
        for iter_time in range(iter):
            x_data = x_train.loc[iter_time]
            y_data = y_train.loc[iter_time]
            x_data = np.mat(x_data.tolist())
            y_data = np.mat(y_data.tolist())
            # 修正参数
            if -1 * y_data * (w * x_data.T + b) >= 0:
                w += learn_rate * y_data * x_data
                b += learn_rate * y_data
        self.w = w
        self.b = b

    def predict(self, x_test: pd.DataFrame):
        count = x_test.shape[0]  # 测试集数量
        w = self.w
        b = self.b
        result = []
        for i in range(count):
            x_data = x_test.loc[i]
            x_data = np.mat(x_data.tolist())
            y = w * x_data.T + b
            result.append(1) if y > 0 else result.append(-1)
        return result


data = pd.read_csv('../Mnist/mnist_train_little.csv', header=None)
test_data = pd.read_csv('../Mnist/mnist_test_little.csv', header=None)
x_train = data.drop(0, axis=1) # 训练集特征
y_train = data[0] # 训练集标签
x_test = test_data.drop(0, axis=1) # 测试集特征

# 模型
perceptron_model = Perceptron()
perceptron_model.fit(x_train=x_train, y_train=y_train)
y_predict = perceptron_model.predict(x_test=x_test)