matlab写粒子群优化算法神经网络

粒子群优化算法（Particle Swarm Optimization，PSO）是一种基于群体智能的优化算法，常用于神经网络的训练中。下面是使用 MATLAB 编写 PSO 神经网络的示例代码：

% 导入数据
load iris_dataset;

% 设置神经网络参数
input_layer_size = 4; % 输入层大小
hidden_layer_size = 5; % 隐藏层大小
num_labels = 3; % 输出层大小
lambda = 1; % 正则化参数

% 初始化神经网络参数
initial_Theta1 = rand(hidden_layer_size, input_layer_size + 1) * 2 - 1;
initial_Theta2 = rand(num_labels, hidden_layer_size + 1) * 2 - 1;
initial_nn_params = [initial_Theta1(:) ; initial_Theta2(:)];

% 定义代价函数
function [J, grad] = nnCostFunction(nn_params, input_layer_size, hidden_layer_size, num_labels, X, y, lambda)
    Theta1 = reshape(nn_params(1:hidden_layer_size * (input_layer_size + 1)), hidden_layer_size, (input_layer_size + 1));
    Theta2 = reshape(nn_params((1 + (hidden_layer_size * (input_layer_size + 1))):end), num_labels, (hidden_layer_size + 1));

    m = size(X, 1);
    J = 0;
    Theta1_grad = zeros(size(Theta1));
    Theta2_grad = zeros(size(Theta2));

    % 前向传播
    a1 = [ones(m,1) X];
    z2 = a1 * Theta1';
    a2 = [ones(size(z2,1),1) sigmoid(z2)];
    z3 = a2 * Theta2';
    h = sigmoid(z3);

    % 计算代价函数
    yVec = zeros(m,num_labels);
    for i=1:num_labels
        yVec(:,i) = (y==i);
        J = J + 1/m * sum(-yVec(:,i) .* log(h(:,i)) - (1-yVec(:,i)) .* log(1-h(:,i)));
    end
    J = J + lambda/(2*m) * (sum(sum(Theta1(:,2:end).^2)) + sum(sum(Theta2(:,2:end).^2)));

    % 反向传播
    delta3 = h - yVec;
    delta2 = delta3 * Theta2(:,2:end) .* sigmoidGradient(z2);
    Delta1 = delta2' * a1;
    Delta2 = delta3' * a2;

    % 计算梯度
    Theta1_grad = Delta1/m + lambda/m * [zeros(size(Theta1,1),1) Theta1(:,2:end)];
    Theta2_grad = Delta2/m + lambda/m * [zeros(size(Theta2,1),1) Theta2(:,2:end)];
    grad = [Theta1_grad(:) ; Theta2_grad(:)];
end

% 使用 PSO 算法进行训练
options = optimoptions('particleswarm','MaxIterations',1000);
costFunction = @(p) nnCostFunction(p, input_layer_size, hidden_layer_size, num_labels, X, y, lambda);
[nn_params, cost] = particleswarm(costFunction, numel(initial_nn_params), [], [], options);

% 获取最优参数
Theta1 = reshape(nn_params(1:hidden_layer_size * (input_layer_size + 1)), hidden_layer_size, (input_layer_size + 1));
Theta2 = reshape(nn_params((1 + (hidden_layer_size * (input_layer_size + 1))):end), num_labels, (hidden_layer_size + 1));

% 使用最优参数进行预测
a1 = [ones(size(X,1),1) X];
z2 = a1 * Theta1';
a2 = [ones(size(z2,1),1) sigmoid(z2)];
z3 = a2 * Theta2';
h = sigmoid(z3);
[~, p] = max(h, [], 2);

% 计算准确率
fprintf('Training Set Accuracy: %f\n', mean(double(p == y)) * 100);

需要注意的是，上述代码中使用了 sigmoid 函数和其梯度函数 sigmoidGradient，因此需要先定义这两个函数：

function g = sigmoid(z)
    g = 1 ./ (1 + exp(-z));
end

function g = sigmoidGradient(z)
    g = sigmoid(z) .* (1 - sigmoid(z));
end

在这个示例中，我们使用了鸢尾花数据集进行训练和测试，但你可以根据自己的需求修改输入层大小、隐藏层大小、输出层大小、正则化参数等参数。