帮我写一段利用粒子群优化BP神经网络的matlab代码，上下限为50和-50，，输入节点数为2，隐含层数为5，输出节点数为1

以下是一个使用粒子群优化算法优化BP神经网络的Matlab代码：

% 设置输入和输出节点数
input_nodes = 2;
hidden_layers = 5;
output_nodes = 1;

% 设置粒子群优化算法的参数
num_particles = 50; % 粒子数量
max_iterations = 100; % 最大迭代次数
w = 0.729; % 惯性权重
c1 = 1.49445; % 个体学习因子
c2 = 1.49445; % 全局学习因子

% 初始化粒子群
particles = rand(num_particles, (input_nodes + 1) * hidden_layers + (hidden_layers + 1) * output_nodes) * 100 - 50;
particle_best_positions = particles;
particle_best_errors = Inf(1, num_particles);
global_best_position = zeros(1, (input_nodes + 1) * hidden_layers + (hidden_layers + 1) * output_nodes);
global_best_error = Inf;

% 训练BP神经网络
for i = 1:max_iterations
% 计算每个粒子的误差
errors = zeros(1, num_particles);
for j = 1:num_particles
weights = reshape(particles(j, :), input_nodes + 1, []);
[error, ~] = train_bp_nn(weights, input_nodes, hidden_layers, output_nodes);
errors(j) = error;

% 更新粒子的最佳位置
if error < particle_best_errors(j)
particle_best_positions(j, :) = particles(j, :);
particle_best_errors(j) = error;
end

% 更新全局最佳位置
if error < global_best_error
global_best_position = particles(j, :);
global_best_error = error;
end
end

% 更新粒子的速度和位置
for j = 1:num_particles
velocity = w * particles(j, :) + c1 * rand(1, size(particles, 2)) .* (particle_best_positions(j, :) - particles(j, :)) + c2 * rand(1, size(particles, 2)) .* (global_best_position - particles(j, :));
particles(j, :) = particles(j, :) + velocity;

% 确保位置在上下限之间
particles(j, :) = max(particles(j, :), -50);
particles(j, :) = min(particles(j, :), 50);
end
end

% 训练后的BP神经网络的权重矩阵
best_weights = reshape(global_best_position, input_nodes + 1, []);

% 函数train_bp_nn的实现
function [error, output] = train_bp_nn(weights, input_nodes, hidden_layers, output_nodes)
% 生成训练数据
X = randn(100, input_nodes);
y = randn(100, output_nodes);

% 设置BP神经网络的参数
learning_rate = 0.1;
max_iterations = 1000;
hidden_layer_sizes = repmat(hidden_layers, 1, 1);

% 初始化BP神经网络
net = feedforwardnet(hidden_layer_sizes, 'traingd');
net.trainParam.lr = learning_rate;
net.trainParam.epochs = max_iterations;
net.trainParam.showWindow = false;
net.trainParam.showCommandLine = false;
net.trainParam.show = NaN;
net.trainParam.goal = 0;
net.trainParam.max_perf_inc = 1.04;
net.trainParam.min_grad = 1e-6;

% 设置权重矩阵
net.IW{1} = weights(1:end-1, 1:hidden_layers);
net.b{1} = weights(end, 1:hidden_layers)';
net.LW{2,1} = weights(1:end-1, hidden_layers+1:end)';
net.b{2} = weights(end, hidden_layers+1:end);

% 训练BP神经网络
net = train(net, X', y');

% 计算误差和输出
output = net(X');
error = mean((output - y').^2);
end