Softmax算法实现

数据集MNIST。

#!/usr/bin/env python3
# -*- coding: utf-8 -*-

import time
import math
import random

import numpy as np
import pandas as pd

from sklearn.model_selection import train_test_split

class Softmax(object):
	def __init__(self):
		self.learning_rate = 0.000001  # 学习速率
		self.max_iteration = 100000  # 最大迭代次数
		self.weight_lambda = 0.01  # 衰退权重

	def cal_e(self, x, l):
		theta_1 = self.w[l]
		product = np.dot(theta_1, x)
		return math.exp(product)

	def cal_probability(self, x, j):
		molecule = self.cal_e(x, j)
		denominator = sum([self.cal_e(x, i) for i in range(self.k)])
		return molecule / denominator

	def cal_partial_derivative(self, x, y, j):
		first = int(y == j)  # 计算I示性函数
		second = self.cal_probability(x, j)  # 计算概率
		# 当j是真实标签类别时，要减去1，否则减去0（结合笔记的公式理解）
		return x * (second - first) + self.weight_lambda * self.w[j]  # 对w的偏导

	def train(self, features, labels):
		self.k = len(set(labels))  # 分类数

		# 权重矩阵，画图理解W*x的过程
		self.w = np.zeros((self.k, len(features[0]) + 1))
		iteration = 0

		while iteration < self.max_iteration:
			# print('loop %d:' % iteration)
			iteration += 1
			# 每次随机选一个点用来计算梯度并迭代
			index = random.randint(0, len(labels) - 1)

			x = features[index]
			y = labels[index]

			x = list(x)
			x.append(1.0)  # 为了与偏置乘
			x = np.array(x)

			derivatives = [self.cal_partial_derivative(x, y, j) for j in range(self.k)]

			for j in range(self.k):
				self.w[j] -= self.learning_rate * derivatives[j]

	def _predict(self, x):
		result = np.dot(self.w, x)
		row, column = result.shape

		# 找出最大值所在列
		_position = np.argmax(result)
		m, _ = divmod(_position, column)

		return m

	def predict(self, features):
		labels = []
		for feature in features:
			x = list(feature)
			x.append(1)

			x = np.matrix(x)
			x = np.transpose(x)

			labels.append(self._predict(x))
		return labels

if __name__ == '__main__':
	print('Start reading data:')
	time1 = time.time()

	raw_data = pd.read_csv('data/train.csv')
	data = raw_data.values

	imgs = data[:, 1:]
	labels = data[:, 0]

	train_features, test_features, train_labels, test_labels = train_test_split(imgs, labels, test_size=0.33, random_state=11111)

	time2 = time.time()
	print('reading data cost %f seconds' % (time2 - time1))

	print('Start training:')
	softmax = Softmax()
	softmax.train(train_features, train_labels)
	time3 = time.time()
	print('training cost %f seconds' % (time3 - time2))

	print('Start predicting:')
	test_predict = softmax.predict(test_features)
	time4 = time.time()
	print('predicting cost %f seconds' % (time4 - time3))

	accuracy = sum([test_labels[i] == test_predict[i] for i in range(len(test_labels))]) / len(test_labels)
	print('The accuracy is %f!' % accuracy)

'''
output:
Start reading data:
reading data cost 5.960523 seconds
Start training:
training cost 39.005752 seconds
Start predicting:
predicting cost 1.706823 seconds
The accuracy is 0.871573!
'''