python卷积神经网络实现

#!/usr/bin/env python3 # -*- coding: utf-8 -*- """ Created on Fri Jan 12 20:30:39 2018 @author: nanzheng """ import numpy as np import matplotlib.pyplot as plt from cnn_operations import cnn_operations as cnn_opr class cnn(): """ 卷积神经网络。 """ def __init__(self): # 网络的层数 self.n_layers = 0 # list，网络中的各层 self.layers = [] # array，网络的输出 self.output = None # 网络的代价 self.loss = None # 权值的学习率 self.learning_rate_weight = 0.1 # 偏置的学习率 self.learning_rate_bias = 0.1 def config(self, args): """ 配置网络。 Parameters ---------- args: tuple，args[i][0]为第i层的类型，{"input", "convoluting", "pooling", "full_connecting", "output"}，args[i][1]为tuple，第i层的配置参数 """ self.n_layers = len(args) prior_layer = None for i in range(self.n_layers): # 配置网络的各层 new_layer = cnn_layer(args[i][0]) if i > 0: prior_layer = self.layers[-1] # 当前层设为上一层的next_layer self.layers[-1].next_layer = new_layer new_layer.config(args[i][1], self.learning_rate_weight, self.learning_rate_bias, prior_layer) self.layers.append(new_layer) return None def _feed_forward(self, x): """ 前向传播。 Parameters ---------- x: 3-d array，一个batch的输入图像， 每个通道的尺寸为x.shape[0] * x.shape[1]， x.shape[2]为当前batch中图像的个数 * 每幅图像的通道数 """ # 输入层前向传播 self.layers[0].feed_forward(x) # 其它各层前向传播 for i in range(1, self.n_layers): self.layers[i].feed_forward(x) # self.layers[-1].n_nodes * size_batch array，网络的输出 self.output = np.ndarray.flatten( \ np.array(self.layers[-1].output)).reshape( \ self.layers[-1].n_nodes, -1) return None def _back_propagate(self, y): """ 反向传播。 Parameters ---------- y: array，输入样本对应的类别标签 """ # 输出层反向传播 self.layers[-1].back_propagate(y) # 其它各层反向传播 for i in range(self.n_layers - 2, 0, -1): self.layers[i].back_propagate() return None def fit(self, X, Y, size_batch=1, n_epochs=1): """ 训练。 下降方式为随机梯度下降。 Parameters ---------- X: 3-d array，训练集， X[:, :, i: i + self.layers[0].n_nodes]为一个训练样本（图片）， self.layers[0].n_nodes即为每幅图片的通道数 Y: array，训练集对应的类别标签 size_batch: 一个batch中训练样本的个数 n_epochs: 迭代次数 """ self.size_batch = size_batch # 训练样本个数 * 每幅图片的通道数 len_X = X.shape[-1] len_Y = Y.shape[0] # 每个epoch中batch的个数 n_batches = int(np.ceil(len_X / self.layers[0].n_nodes / size_batch)) loss = np.empty(n_epochs * n_batches) for i_epoch in range(n_epochs): print("Epoch: ", end="") print(i_epoch) for i_batch in range(n_batches): print("\tBatch: ", end="") print(i_batch, end="\t") y_offset = i_batch * size_batch x_offset = y_offset * self.layers[0].n_nodes # 将类别标签转换为向量 y = np.zeros([self.layers[-1].n_nodes, size_batch]) for i in range(size_batch): if i > len_Y - y_offset - 1: y = y[:, :, : i] break y[Y[y_offset + i], i] = 1 self._feed_forward(X[:, :, x_offset: x_offset + size_batch * \ self.layers[0].n_nodes]) loss[i_epoch * n_batches + i_batch] = \ cnn_opr.calc_loss(y.T, self.output.T) print("loss = ", end="") print(loss[i_epoch * n_batches + i_batch]) self._back_propagate(y) self.loss = loss plt.figure() plt.plot(loss, "r-") plt.xlabel("Batches") plt.ylabel("Loss") plt.grid() plt.show() return None def test(self, X, Y): """ 验证。 Parameters ---------- X: 3-d array，验证集， X[:, :, i: i + self.layers[0].n_nodes]为一个验证样本（图片）， self.layers[0].n_nodes即为每幅图片的通道数 Y: array，训练集对应的类别标签 Returns ------- correct_rate: 验证集的分类正确率 """ n_correct = 0 for i in range(0, X.shape[-1], self.layers[0].n_nodes): print("Test case: ", end="") print(i) y_predict = self.predict(X[:, :, i: i + self.layers[0].n_nodes]) if y_predict == Y[i]: n_correct += 1 correct_rate = n_correct / X.shape[-1] return correct_rate def predict(self, x): """ 预测。 Parameters ---------- x: 2-d或3-d array，输入样本（图像） Returns ------- y_predict: 预测出的输入样本（图像）的类别 """ self._feed_forward(x.reshape(x.shape[0], x.shape[1], -1)) # 根据网络输出层的类型，判定输入图像的类别 if self.layers[-1].type_output is "softmax": y_predict = np.argmax(self.output[:, 0]) elif self.layers[-1].type_output is "rbf": # TODO: pass return y_predict class cnn_layer(): """ 卷积神经网络中的一层。 """ def __init__(self, type_layer): """ Parameters ---------- type_layer: 当前层的类型，{"input", "convoluting", "pooling", "full_connecting", "output"} """ # 当前层的类型 self.type = type_layer # 当前层中神经元的个数 self.n_nodes = 0 # list，当前层中各神经元 self.nodes = [] # 当前层的上一层 self.prior_layer = None # 当前层的下一层 self.next_layer = None # list，当前层的输出 self.output = [] # 权值的学习率 self.learning_rate_weight = 0.0 # 偏置的学习率 self.learning_rate_bias = 0.0 if self.type is "input": # array，输入图像（每个通道）的尺寸 self.size_input = None elif self.type is "convoluting": # 2-d array，当前层与上一层各神经元的连接矩阵 self.connecting_matrix = None # array，卷积核尺寸 self.size_conv_kernel = None