python3中的zip函数

python2中（

python3中（https://docs.python.org/3.3/library/functions.html#zip）

背景:

不经常写python的我，今天在读https://ff120.github.io/2017/05/19/%E6%9C%BA%E5%99%A8%E5%AD%A6%E4%B9%A0%E4%B8%93%E9%A2%98/%E6%9C%BA%E5%99%A8%E5%AD%A6%E4%B9%A0_%E6%84%9F%E7%9F%A5%E6%9C%BA/

这篇博客时,把文章中的python2语法改成python3时遇到了一个让我感觉到有些意外的情况;

先放一下对应python3下的代码:

from functools import reduce
from itertools import tee

class Perceptron:
	def __init__(self,input_num,activator):
		self.activator = activator
		self.weights = [0.0 for _ in range(input_num)]
		self.bias = 0.0
	def __str__(self):
		return 'weights\t:%s\nbias\t:%f\n' % (self.weights, self.bias)
	def predict(self, input_vec):
		return self.activator(\
			reduce(lambda a,b:a+b,\
				map(lambda x_w:x_w[0]*x_w[1], \
					zip(input_vec,self.weights)) \
					,0.0)+self.bias)

	def _update_weight(self,input_vec,output,label,rate):
		delta = label - output
		self.weights = list(map(
		lambda x_w:x_w[1]+rate*delta*x_w[0],
		zip(input_vec,self.weights)))
		self.bias += rate * delta

	def _one_iteration(self,input_vecs,labels,rate):
		samples = zip(input_vecs,labels)
		#print(list(samples))
		for(input_vec,label) in samples:
			output = self.predict(input_vec)
			self._update_weight(input_vec,output,label,rate)

	def train(self,input_vecs,labels,iteration,rate):
		for i in range(iteration):
			self._one_iteration(input_vecs,labels,rate)


def and_f(x):
	return 1 if x>0 else 0

if __name__ == '__main__':
	x = [[0,0],[0,1],[1,0],[1,1]]
	y = [0,0,0,1]
	and_perception = Perceptron(2,and_f)
	and_perception.train(x,y,10,0.1)
	print(and_perception)
	print('1 and 1 = %d'%and_perception.predict([1,1]))
	print('0 and 0 = %d'%and_perception.predict([0,0]))
	print('1 and 0 = %d'%and_perception.predict([1,0]))
	print('0 and 1 = %d'%and_perception.predict([0,1]))		

现象就是当我将注释的print(list(samples))打开时会影响到后续的结果,感觉就像list类型强转改变了samples一样;

查了查原来问题出现在zip上,python3的zip返回迭代器(从性能方面看这样也许确实更好吧);

当修改成

	samples = list(zip(input_vecs,labels))
	print(list(samples))

或者

	samples = zip(input_vecs,labels)
	samples,second_it = tee(samples)
	print(list(samples))
	samples = second_it

就好了;

同样可以注意到python3 map返回也是个迭代器;

参考:

https://stackoverflow.com/questions/8689184/nameerror-name-reduce-is-not-defined-in-python

https://stackoverflow.com/questions/40912089/python-map-on-zipped-object

https://stackoverflow.com/questions/21892989/what-is-the-good-python3-equivalent-for-auto-tuple-unpacking-in-lambda

https://stackoverflow.com/questions/50671360/map-in-python-3-vs-python-2

https://stackoverflow.com/questions/19777612/python-range-and-zip-object-type

https://stackoverflow.com/questions/42132731/how-to-create-a-copy-of-python-iterator

https://stackoverflow.com/questions/31683959/the-zip-function-in-python-3

https://blog.csdn.net/benpaodelulu_guajian/article/details/81869462

/*************************************************************************************************/

下面我们对上面的python感知机作一些可视化

from matplotlib import pyplot
from matplotlib import cm
from mpl_toolkits.mplot3d import Axes3D
import numpy as np

if __name__ == '__main__':
	# 训练数据
	x = np.array([[0,0,0],[1,1,1],[1,1,0],[1,0,1],[0,1,1]])
	y = [1,0,0,0,0]
	and_perception = Perceptron(3,and_f)
	and_perception.train(x,y,10,0.1)
	print(and_perception)
	fig = pyplot.figure()
	pyplot.xlim(0, 1)
	pyplot.ylim(0, 1)
	ax = Axes3D(fig)
	ax.set_zlim3d(0,1)
	length = len(y)
	for i in range(length):
		if y[i]==1:
			ax.scatter(x[0:,0:1][i],x[0:,1:2][i],x[0:,2:3][i], c = 'r', marker='*')
		else:
			ax.scatter(x[0:,0:1][i],x[0:,1:2][i],x[0:,2:3][i], c = 'b', marker='*')

	x2 = np.linspace(0,1,10)
	y2 = np.linspace(0,1,10)
	X,Y = np.meshgrid(x2,y2)
	Z = (and_perception.weights[0]*X + and_perception.weights[1]*Y +  + and_perception.bias)/(-and_perception.weights[2])
	for idx1, zz in enumerate(Z):
		for idx2, zzz in enumerate(zz):
			if Z[idx1][idx2]<0 :			
				Z[idx1][idx2] = and_perception.bias/(-and_perception.weights[2])#Z[idx1][idx2] if Z[idx1][idx2]>0 else np.nan
				X[idx1][idx2] = 0
				Y[idx1][idx2] = 0
	surf = ax.plot_surface(X, Y, Z, rstride = 1, cstride = 1,cmap = 'viridis', edgecolor = 'none',vmin=np.nanmin(Z), vmax=np.nanmax(Z))
	#使用模型对100个随机三维点进行分类
	test = np.random.randint(0,1000,(100,3))/1000
	for t in test:
		if and_perception.predict(t)==1:
			ax.scatter(t[0],t[1],t[2], c = 'r', marker='*')
		else:
			ax.scatter(t[0],t[1],t[2], c = 'b', marker='*')

	ax.set_xlabel('X-axis')
	ax.set_ylabel('Y-axis')
	ax.set_zlabel('Z-axis')
	pyplot.show()

  参考:

单层感知器算法（可用于与或非）

感知机搞不定逻辑XOR？Science新研究表示人脑单个神经元就能做到

机器学习之神经网络（上）

简单易懂的讲解深度学习（入门系列之五）

Calculate the Decision Boundary of a Single Perceptron - Visualizing Linear Separability

Perceptron in Python

Python Machine Learning Tutorial

Perceptron

[資料分析&機器學習] 第3.2講：線性分類-感知器(Perceptron) 介紹

Perceptron Learning Algorithm: A Graphical Explanation Of Why It Works

【ML学习笔记】18：原始的Perceptron(感知机)

python matplotlib 2D 散点图的制作（输入二维坐标）

how to do the scatter plot for the lists or 2d array or matrix python?

scatterplotmatrix.py

Iterate over a list in Python

Plot 3D mesh using mplot3d

Python三维绘图--Matplotlib

python 画直线和平面

python - Python-Matplotlib-如何从方程式绘制平面？

有效可视化多维数据的艺术 有效数据可视化的策略

Matplotlib中的三维绘图

Matplotlib学习笔记——画三维图

坐标轴范围的设置

Set zlim in matplotlib scatter3d

Python - Set zlim in mplot3D

.pyplot has not attribute zlim

Python/matplotlib mplot3d- how do I set a maximum value for the z-axis?

Matplotlib数据分析基础课程,三剑客,之,基础教程

Getting the class name of an instance?

Most efficient way to map function over numpy array

python中的三元表达式（三目运算符）

Iterate and modify array with NumPy

Modify NumPy array in loops

Python：有没有办法用Matplotlib绘制“局部”表面图？

Simple way to create matrix of random numbers

Python time.sleep() 阻塞指定秒杀 random.randint()返回指定闭区间内的随机数

numpy.random模块用法总结

https://github.com/cezannec/intro-computervision intro-computervision

https://cezannec.github.io/Convolutional_Neural_Networks/ 卷积神经网络

https://github.com/ahmedfgad/NumPyCNN NumPyCNN

https://stackoverflow.com/questions/55807920/python-how-to-install-scikit-image-when-the-installation-appears-broken Python, how to install scikit-image when the installation appears broken

https://stackoverflow.com/questions/14812342/matplotlib-has-no-attribute-pyplot matplotlib-has-no-attribute-pyplot

https://towardsdatascience.com/convolutional-neural-networks-for-beginners-practical-guide-with-python-and-keras-dc688ea90dca 初学者的卷积神经网络

https://towardsdatascience.com/building-a-convolutional-neural-network-cnn-in-keras-329fbbadc5f5 Building a Convolutional Neural Network (CNN) in Keras

https://www.datacamp.com/community/tutorials/convolutional-neural-networks-python Convolutional Neural Networks in Python with Keras

https://www.tensorflow.org/tutorials/images/cnn Convolutional Neural Network (CNN)

https://missinglink.ai/guides/convolutional-neural-networks/python-convolutional-neural-network-creating-cnn-keras-tensorflow-plain-python/ Python卷积神经网络：在Keras，TensorFlow和Plain Python中创建CNN

https://www.kdnuggets.com/2018/04/building-convolutional-neural-network-numpy-scratch.html Building Convolutional Neural Network using NumPy from Scratch

http://www.skcircle.com/?id=329 滤波的概念和作用（滤波器、掩模、核、模板、窗口是一个意思）

https://cloud.tencent.com/developer/article/1198216 卷积滤波器与边缘检测

https://cloud.tencent.com/developer/article/1600526 数字图像处理中常见的变换核及其用途

https://zhuanlan.zhihu.com/p/45585819 opencv处理图像

https://zhuanlan.zhihu.com/p/81792708 卷积的通俗理解——从傅里叶变换到滤波器

https://opencv-python-tutorials.readthedocs.io/zh/latest/4.%20OpenCV%E4%B8%AD%E7%9A%84%E5%9B%BE%E5%83%8F%E5%A4%84%E7%90%86/4.4.%20%E5%9B%BE%E5%83%8F%E6%BB%A4%E6%B3%A2/ 2D卷积（图像过滤）

https://www.cnblogs.com/carsonzhu/p/10831624.html 计算机视觉】OpenCV篇(6) - 平滑图像（卷积/滤波/模糊/降噪）