可变多隐层神经网络的python实现

说明：这是我对网上代码的改写版本，目的是使它跟前一篇提到的使用方法尽量一致，用起来更直观些。

此神经网络有两个特点：

1、灵活性

非常灵活，隐藏层的数目是可以设置的，隐藏层的激活函数也是可以设置的

2、扩展性

扩展性非常好。目前只实现了一个学习方法：lm（Levenberg-Marquardt训练算法），你可以添加不同的学习方法到NeuralNetwork类

      什么是最优化，可分为几大类？
    答：Levenberg-Marquardt算法是最优化算法中的一种。最优化是寻找使得函数值最小的参数向量。它的应用领域非常广泛，如：经济学、管理优化、网络分析、最优设计、机械或电子设计等等。
    根据求导数的方法，可分为2大类。第一类，若f具有解析函数形式，知道x后求导数速度快。第二类，使用数值差分来求导数。
    根据使用模型不同，分为非约束最优化、约束最优化、最小二乘最优化。

      什么是Levenberg-Marquardt算法？
    答：它是使用最广泛的非线性最小二乘算法，中文为列文伯格-马夸尔特法。它是利用梯度求最大（小）值的算法，形象的说，属于“爬山”法的一种。它同时具有梯度法和牛顿法的优点。当λ很小时，步长等于牛顿法步长，当λ很大时，步长约等于梯度下降法的步长。

 （本文基于win7 + python3.4）

一、先说说改动的部分

1) 改写了NetStruct类初始化函数

原来：

class NetStruct: 
    '''神经网络结构'''
    def __init__(self, x, y, hidden_layers, activ_fun_list, performance_function = 'mse'):

现在：

class NetStruct: 
    '''神经网络结构'''
    def __init__(self, ni, nh, no, active_fun_list):
        # ni 输入层节点数（int）
        # ni 隐藏层节点数（int 或 list）
        # no 输出层节点数（int）
        # active_fun_list 隐藏层激活函数类型（list）

好处：

      初始化网络时更直观

2) 修改了NeuralNetwork类的train函数的参数

原来：

class NeuralNetwork:

    def __init__(self, ...):
        # ...

    def train(self, method = 'lm'):
        if(method == 'lm'):
            self.lm()

现在：

class NeuralNetwork:
    '''神经网络'''
    def __init__(self,  ...):
        '''初始化'''
        # ...
    
    def train(self, x, y, method = 'lm'):
        '''训练'''
        self.net_struct.x = x
        self.net_struct.y = y
        if(method == 'lm'):
            self.lm()

好处：

      使用训练（train）函数时更直观

3) 修改了sinSamples样例函数的错误

　　原代码会报错："x与y维度不一致"

4) 添加了部分注释

      主要是为了更好地理解代码

二、再说说隐藏层的层数对误差的影响

效果图（peakSamples样例）

1) 两个隐藏层

误差：

100次迭代后，误差一般收敛在（1.3， 0.3）这个区间

2) 三个隐藏层

误差：

100次迭代后，误差一般收敛在（0.3， 0.1）这个区间

 结论：对于peakSamples这个样例，采用三个隐藏层优于两个隐藏层（显然，这里没有考虑与测试各隐藏层神经元数目改变的情况）

注：还可以试试另外一个测试函数

再附一个我用股票数据做的预测效果图 ：

完整代码：

# neuralnetwork.py
# modified by Robin 2015/03/03

import numpy as np
from math import exp, pow
from mpl_toolkits.mplot3d import Axes3D
import matplotlib.pyplot as plt
import sys
import copy
from scipy.linalg import norm, pinv

class Layer: 
    '''层'''
    def __init__(self, w, b, neure_number, transfer_function, layer_index):
        self.transfer_function = transfer_function
        self.neure_number = neure_number
        self.layer_index = layer_index
        self.w = w
        self.b = b

class NetStruct: 
    '''神经网络结构'''
    def __init__(self, ni, nh, no, active_fun_list):
        # ni 输入层节点（int）
        # ni 隐藏层节点（int 或 list）
        # no 输出层节点（int）
        # active_fun_list 隐藏层激活函数类型（list）
        # ==> 1
        self.neurals = [] # 各层的神经元数目
        self.neurals.append(ni)
        if isinstance(nh, list):
            self.neurals.extend(nh)
        else:
            self.neurals.append(nh)
        self.neurals.append(no)
        # ==> 2
        if len(self.neurals)-2 == len(active_fun_list):
            active_fun_list.append('line')
        self.active_fun_list = active_fun_list
        # ==> 3
        self.layers = [] # 所有的层
        for i in range(0, len(self.neurals)):
            if i == 0:
                self.layers.append(Layer([], [], self.neurals[i], 'none', i)) 
                continue
            f = self.neurals[i - 1]
            s = self.neurals[i] 
            self.layers.append(Layer(np.random.randn(s, f), np.random.randn(s, 1), self.neurals[i], self.active_fun_list[i-1], i))
        
class NeuralNetwork:
    '''神经网络'''
    def __init__(self, net_struct, mu = 1e-3, beta = 10, iteration = 100, tol = 0.1):
        '''初始化'''
        self.net_struct = net_struct
        self.mu = mu
        self.beta = beta
        self.iteration = iteration
        self.tol = tol
    
    def train(self, x, y, method = 'lm'):
        '''训练'''
        self.net_struct.x = x
        self.net_struct.y = y
        if(method == 'lm'):
            self.lm()
    
    def sim(self, x):
        '''预测'''
        self.net_struct.x = x
        self.forward()
        layer_num = len(self.net_struct.layers)
        predict = self.net_struct.layers[layer_num - 1].output_val
        return predict
    
    def actFun(self, z, active_type = 'sigm'):
        ''' 激活函数 '''
        # activ_type: 激活函数类型有 sigm、tanh、radb、line
        if active_type == 'sigm':
            f = 1.0 / (1.0 + np.exp(-z))
        elif active_type == 'tanh':
            f = (np.exp(z) + np.exp(-z)) / (np.exp(z) + np.exp(-z))
        elif active_type == 'radb':
            f = np.exp(-z * z)
        elif active_type == 'line':
            f = z
        return f
    
    def actFunGrad(self, z, active_type = 'sigm'):
        '''激活函数的变化（派生）率'''
        # active_type: 激活函数类型有 sigm、tanh、radb、line
        y = self.actFun(z, active_type)
        if active_type == 'sigm':
            grad = y * (1.0 - y)
        elif active_type == 'tanh':
            grad = 1.0 - y * y
        elif active_type == 'radb':
            grad = -2.0 * z * y
        elif active_type == 'line':
            m = z.shape[0]
            n = z.shape[1]
            grad = np.ones((m, n))
        return grad
    
    def forward(self): 
        ''' 前向 '''
        layer_num = len(self.net_struct.layers)
        for i in range(0, layer_num):
            if i == 0:
                curr_layer = self.net_struct.layers[i]
                curr_layer.input_val = self.net_struct.x
                curr_layer.output_val = self.net_struct.x
                continue
            before_layer = self.net_struct.layers[i - 1]
            curr_layer = self.net_struct.layers[i]
            curr_layer.input_val = curr_layer.w.dot(before_layer.output_val) + curr_layer.b
            curr_layer.output_val = self.actFun(curr_layer.input_val, 
                                                self.net_struct.active_fun_list[i - 1])
    
    def backward(self):
        '''反向'''
        layer_num = len(self.net_struct.layers)
        last_layer = self.net_struct.layers[layer_num - 1]
        last_layer.error = -self.actFunGrad(last_layer.input_val,
                                            self.net_struct.active_fun_list[layer_num - 2])
        layer_index = list(range(1, layer_num - 1)) 
        layer_index.reverse()
        for i in layer_index:
            curr_layer = self.net_struct.layers[i]
            curr_layer.error = (last_layer.w.transpose().dot(last_layer.error)) * self.actFunGrad(curr_layer.input_val,self.net_struct.active_fun_list[i - 1])
            last_layer = curr_layer
    
    def parDeriv(self):
        '''标准梯度（求导）'''
        layer_num = len(self.net_struct.layers)
        for i in range(1, layer_num):
            befor_layer = self.net_struct.layers[i - 1]
            befor_input_val = befor_layer.output_val.transpose()
            curr_layer = self.net_struct.layers[i]
            curr_error = curr_layer.error
            curr_error = curr_error.reshape(curr_error.shape[0]*curr_error.shape[1], 1, order='F')
            row =  curr_error.shape[0]
            col = befor_input_val.shape[1]
            a = np.zeros((row, col))
            num = befor_input_val.shape[0]
            neure_number = curr_layer.neure_number
            for i in range(0, num):
                a[neure_number*i:neure_number*i + neure_number,:] = np.repeat([befor_input_val[i,:]],neure_number,axis = 0)
            tmp_w_par_deriv = curr_error * a
            curr_layer.w_par_deriv = np.zeros((num, befor_layer.neure_number * curr_layer.neure_number))
            for i in range(0, num):
                tmp = tmp_w_par_deriv[neure_number*i:neure_number*i + neure_number,:]
                tmp = tmp.reshape(tmp.shape[0] * tmp.shape[1], order='C')
                curr_layer.w_par_deriv[i, :] = tmp
                curr_layer.b_par_deriv = curr_layer.error.transpose()
    
    def jacobian(self):
        '''雅可比行列式'''
        layers = self.net_struct.neurals
        row = self.net_struct.x.shape[1]
        col = 0
        for i in range(0, len(layers) - 1):
            col = col + layers[i] * layers[i + 1] + layers[i + 1]
        j = np.zeros((row, col))
        layer_num = len(self.net_struct.layers)
        index = 0
        for i in range(1, layer_num):
            curr_layer = self.net_struct.layers[i]
            w_col = curr_layer.w_par_deriv.shape[1]
            b_col = curr_layer.b_par_deriv.shape[1]
            j[:, index : index + w_col] = curr_layer.w_par_deriv
            index = index + w_col
            j[:, index : index + b_col] = curr_layer.b_par_deriv
            index = index + b_col
        return j
    
    def gradCheck(self):
        '''梯度检查'''
        W1 = self.net_struct.layers[1].w
        b1 = self.net_struct.layers[1].b
        n = self.net_struct.layers[1].neure_number
        W2 = self.net_struct.layers[2].w
        b2 = self.net_struct.layers[2].b
        x = self.net_struct.x
        p = []
        p.extend(W1.reshape(1,W1.shape[0]*W1.shape[1],order = 'C')[0])
        p.extend(b1.reshape(1,b1.shape[0]*b1.shape[1],order = 'C')[0])
        p.extend(W2.reshape(1,W2.shape[0]*W2.shape[1],order = 'C')[0])
        p.extend(b2.reshape(1,b2.shape[0]*b2.shape[1],order = 'C')[0])
        old_p = p
        jac = []
        for i in range(0, x.shape[1]):
            xi = np.array([x[:,i]])
            xi = xi.transpose()
            ji = []
            for j in range(0, len(p)):
                W1 = np.array(p[0:2*n]).reshape(n,2,order='C')
                b1 = np.array(p[2*n:2*n+n]).reshape(n,1,order='C')
                W2 = np.array(p[3*n:4*n]).reshape(1,n,order='C')
                b2 = np.array(p[4*n:4*n+1]).reshape(1,1,order='C')

                z2 = W1.dot(xi) + b1
                a2 = self.actFun(z2)
                z3 = W2.dot(a2) + b2
                h1 = self.actFun(z3)
                p[j] = p[j] + 0.00001
                W1 = np.array(p[0:2*n]).reshape(n,2,order='C')
                b1 = np.array(p[2*n:2*n+n]).reshape(n,1,order='C')
                W2 = np.array(p[3*n:4*n]).reshape(1,n,order='C')
                b2 = np.array(p[4*n:4*n+1]).reshape(1,1,order='C')

                z2 = W1.dot(xi) + b1
                a2 = self.actFun(z2)
                z3 = W2.dot(a2) + b2
                h = self.actFun(z3)
                g = (h[0][0]-h1[0][0])/0.00001
                ji.append(g)
            jac.append(ji)
            p = old_p
        return jac
    
    def jjje(self):
        '''计算jj与je'''
        layer_num = len(self.net_struct.layers)
        e = self.net_struct.y - self.net_struct.layers[layer_num - 1].output_val
        e = e.transpose()
        j = self.jacobian()
        #check gradient
        #j1 = -np.array(self.gradCheck())
        #jk = j.reshape(1,j.shape[0]*j.shape[1])
        #jk1 = j1.reshape(1,j1.shape[0]*j1.shape[1])
        #plt.plot(jk[0])
        #plt.plot(jk1[0],'.')
        #plt.show()
        jj = j.transpose().dot(j)
        je = -j.transpose().dot(e)
        return[jj, je]
    
    def lm(self):
        '''Levenberg-Marquardt训练算法'''
        mu = self.mu
        beta = self.beta
        iteration = self.iteration
        tol = self.tol
        y = self.net_struct.y
        layer_num = len(self.net_struct.layers)
        self.forward()
        pred =  self.net_struct.layers[layer_num - 1].output_val
        pref = self.perfermance(y, pred)
        for i in range(0, iteration):
            print('iter:',i, 'error:', pref)
            #1) 第一步: 
            if(pref < tol):
                break
            #2) 第二步:
            self.backward()
            self.parDeriv()
            [jj, je] = self.jjje()
            while(1):
                #3) 第三步: 
                A = jj + mu * np.diag(np.ones(jj.shape[0]))
                delta_w_b = pinv(A).dot(je)
                #4) 第四步:
                old_net_struct = copy.deepcopy(self.net_struct)
                self.updataNetStruct(delta_w_b)
                self.forward()
                pred1 =  self.net_struct.layers[layer_num - 1].output_val
                pref1 = self.perfermance(y, pred1)
                if (pref1 < pref):
                    mu = mu / beta
                    pref = pref1
                    break
                mu = mu * beta
                self.net_struct = copy.deepcopy(old_net_struct)
    
    def updataNetStruct(self, delta_w_b):
        '''更新网络权重及阈值'''
        layer_num = len(self.net_struct.layers)
        index = 0
        for i in range(1, layer_num):
            before_layer = self.net_struct.layers[i - 1]
            curr_layer = self.net_struct.layers[i]
            w_num = before_layer.neure_number * curr_layer.neure_number
            b_num = curr_layer.neure_number
            w = delta_w_b[index : index + w_num]
            w = w.reshape(curr_layer.neure_number, before_layer.neure_number, order='C')
            index = index + w_num
            b = delta_w_b[index : index + b_num]
            index = index + b_num
            curr_layer.w += w
            curr_layer.b += b
    
    def perfermance(self, y, pred):
        '''性能函数'''
        error = y - pred
        return norm(error) / len(y)

        
        
# 以下函数为测试样例      
def plotSamples(n = 40):
    x = np.array([np.linspace(0, 3, n)])
    x = x.repeat(n, axis = 0)
    y = x.transpose()
    z = np.zeros((n, n))
    for i in range(0, x.shape[0]):
        for j in range(0, x.shape[1]):
            z[i][j] = sampleFun(x[i][j], y[i][j])
    fig = plt.figure()
    ax = fig.gca(projection='3d')
    surf = ax.plot_surface(x, y, z, cmap='autumn', cstride=2, rstride=2)
    ax.set_xlabel("X-Label")
    ax.set_ylabel("Y-Label")
    ax.set_zlabel("Z-Label")
    plt.show()

def sinSamples(n):
    x = np.array([np.linspace(-0.5, 0.5, n)])
    #x = x.repeat(n, axis = 0)
    y = x + 0.2
    z = np.zeros((n, 1))
    for i in range(0, x.shape[1]):
        z[i] = np.sin(x[0][i] * y[0][i])
    X = np.zeros((n, 2))
    n = 0
    for xi, yi in zip(x.transpose(), y.transpose()):
        X[n][0] = xi
        X[n][1] = yi
        n = n + 1
    # print(x.shape, y.shape)
    # print(X.shape, z.shape)
    return X, z.transpose()

def peaksSamples(n):
    x = np.array([np.linspace(-3, 3, n)])
    x = x.repeat(n, axis = 0)
    y = x.transpose()
    z = np.zeros((n, n))
    for i in range(0, x.shape[0]):
        for j in range(0, x.shape[1]):
            z[i][j] = sampleFun(x[i][j], y[i][j])
    X = np.zeros((n*n, 2))
    x_list = x.reshape(n*n,1 )
    y_list = y.reshape(n*n,1)
    z_list = z.reshape(n*n,1)
    n = 0
    for xi, yi in zip(x_list, y_list):
        X[n][0] = xi
        X[n][1] = yi
        n = n + 1
    # print(x.shape, y.shape)
    # print(X.shape, z.shape, z_list.shape, z_list.transpose().shape)
    return X,z_list.transpose()

def sampleFun( x, y):
    z =  3*pow((1-x),2) * exp(-(pow(x,2)) - pow((y+1),2))  - 10*(x/5 - pow(x, 3) - pow(y, 5)) * exp(-pow(x, 2) - pow(y, 2)) - 1/3*exp(-pow((x+1), 2) - pow(y, 2)) 
    return z




# 测试
if __name__ == '__main__':

    active_fun_list = ['sigm','sigm','sigm']# 【必须】设置【各】隐层的激活函数类型，可以设置为tanh,radb，tanh,line类型，如果不显式的设置最后一层为line 
    ns = NetStruct(2, [10, 10, 10], 1, active_fun_list) # 确定神经网络结构，中间两个隐层各10个神经元
    nn = NeuralNetwork(ns) # 神经网络类实例化
    
    [X, z] = peaksSamples(20) # 产生训练数据
    #[X, z] = sinSamples(20)  # 第二个训练数据
    X = X.transpose()
    
    # 注意：测试数据的格式与我们习惯的用法有差别，行列要转置！！原因是内部逻辑采用了矩阵运算？！！！！
    #print(X.shape) # (2, 20)
    #print(X)
    #print(z.shape) # (1, 20)
    #print(z)
    
    nn.train(X, z) # 训练！！！！！！
    
    [X0, z0] = peaksSamples(40) # 产生测试数据
    #[X0, z0] = sinSamples(40)  # 第二个测试数据
    X0 = X0.transpose()
    
    z1 = nn.sim(X0) # 预测！！！！！！

    fig  = plt.figure()
    ax = fig.add_subplot(111)
    ax.plot(z0[0])      # 画出真实值 real data
    ax.plot(z1[0],'r.') # 画出预测值 predict data
    plt.legend(('real data', 'predict data'))
    plt.show()