粒子群优化bp神经网络python代码

以下是使用粒子群优化算法优化BP神经网络的Python代码示例：

首先，需要导入需要的库，包括numpy、random和math：

import numpy as np
import random
import math

接下来，定义BP神经网络的类：

class BPNN:
    def __init__(self, input_dim, hidden_dim, output_dim):
        self.input_dim = input_dim
        self.hidden_dim = hidden_dim
        self.output_dim = output_dim
        self.W1 = np.random.randn(self.input_dim, self.hidden_dim)
        self.W2 = np.random.randn(self.hidden_dim, self.output_dim)

    def sigmoid(self, x):
        return 1 / (1 + np.exp(-x))

    def forward(self, X):
        self.z2 = np.dot(X, self.W1)
        self.a2 = self.sigmoid(self.z2)
        self.z3 = np.dot(self.a2, self.W2)
        y_hat = self.sigmoid(self.z3)
        return y_hat

    def cost_func(self, X, y):
        self.y_hat = self.forward(X)
        J = 0.5 * sum((y - self.y_hat) ** 2)
        return J

    def backprop(self, X, y):
        delta3 = np.multiply(-(y - self.y_hat), self.sigmoid(self.z3) * (1 - self.sigmoid(self.z3)))
        dJdW2 = np.dot(self.a2.T, delta3)

        delta2 = np.dot(delta3, self.W2.T) * self.sigmoid(self.z2) * (1 - self.sigmoid(self.z2))
        dJdW1 = np.dot(X.T, delta2)

        return dJdW1, dJdW2

    def predict(self, X):
        y_hat = self.forward(X)
        return y_hat

接下来，定义粒子群优化算法的类：

class PSO:
    def __init__(self, n_particles, n_iterations, c1, c2, w, lr):
        self.n_particles = n_particles
        self.n_iterations = n_iterations
        self.c1 = c1
        self.c2 = c2
        self.w = w
        self.lr = lr

    def optimize(self, bpnn, X, y):
        position = np.random.randn(self.n_particles, bpnn.W1.size + bpnn.W2.size)
        velocity = np.random.randn(self.n_particles, bpnn.W1.size + bpnn.W2.size)
        pbest_position = position.copy()
        pbest_cost = np.zeros(self.n_particles)
        gbest_position = np.zeros(bpnn.W1.size + bpnn.W2.size)
        gbest_cost = float('inf')

        for i in range(self.n_particles):
            bpnn.W1 = position[i][:bpnn.W1.size].reshape(bpnn.W1.shape)
            bpnn.W2 = position[i][bpnn.W1.size:].reshape(bpnn.W2.shape)
            pbest_cost[i] = bpnn.cost_func(X, y)

            if pbest_cost[i] < gbest_cost:
                gbest_cost = pbest_cost[i]
                gbest_position = position[i]

        for i in range(self.n_iterations):
            for j in range(self.n_particles):
                bpnn.W1 = position[j][:bpnn.W1.size].reshape(bpnn.W1.shape)
                bpnn.W2 = position[j][bpnn.W1.size:].reshape(bpnn.W2.shape)

                cost = bpnn.cost_func(X, y)
                if cost < pbest_cost[j]:
                    pbest_cost[j] = cost
                    pbest_position[j] = position[j]

                    if pbest_cost[j] < gbest_cost:
                        gbest_cost = pbest_cost[j]
                        gbest_position = pbest_position[j]

                v = velocity[j]
                r1 = np.random.rand(bpnn.W1.size + bpnn.W2.size)
                r2 = np.random.rand(bpnn.W1.size + bpnn.W2.size)
                v = self.w * v + self.c1 * r1 * (pbest_position[j] - position[j]) + self.c2 * r2 * (gbest_position - position[j])
                position[j] += self.lr * v

        bpnn.W1 = gbest_position[:bpnn.W1.size].reshape(bpnn.W1.shape)
        bpnn.W2 = gbest_position[bpnn.W1.size:].reshape(bpnn.W2.shape)

        return bpnn

最后，使用以上定义的类和函数来训练BP神经网络：

X = np.array([[0, 0], [0, 1], [1, 0], [1, 1]])
y = np.array([[0], [1], [1], [0]])

bpnn = BPNN(2, 3, 1)
pso = PSO(10, 100, 1.5, 1.5, 0.7, 0.1)
bpnn = pso.optimize(bpnn, X, y)

print(bpnn.predict(X))

以上代码实现了一个简单的XOR问题的BP神经网络，并使用粒子群优化算法进行训练。