用MATLAB写一段用训练集和测试集的dnn代码

好的，这是一个简单的用训练集和测试集的DNN代码示例，使用MATLAB的深度学习工具箱：

首先，加载数据集，并将其分成训练集和测试集：

load('dataset.mat'); % 加载数据集
num_samples = size(dataset,1);
num_train = floor(0.8*num_samples); % 80%的数据作为训练集
indices = randperm(num_samples);
train_indices = indices(1:num_train);
test_indices = indices(num_train+1:end);

train_data = dataset(train_indices, 1:end-1);
train_labels = dataset(train_indices, end);
test_data = dataset(test_indices, 1:end-1);
test_labels = dataset(test_indices, end);

接下来，定义DNN的网络结构：

layers = [
    fullyConnectedLayer(50)
    reluLayer
    fullyConnectedLayer(25)
    reluLayer
    fullyConnectedLayer(10)
    softmaxLayer
    classificationLayer
];

这里定义了3个全连接层和2个激活函数层，最后是一个softmax分类层和一个分类层。

然后，定义训练选项和训练DNN：

options = trainingOptions('adam', ...
    'MaxEpochs', 100, ...
    'MiniBatchSize', 32, ...
    'Shuffle', 'every-epoch', ...
    'Plots', 'training-progress');

net = trainNetwork(train_data', categorical(train_labels), layers, options);

这里使用ADAM优化器，最大训练轮数为100，每批训练32个样本，每轮训练后打乱数据集。训练完成后，我们可以使用测试集来评估DNN的性能：

YTest = classify(net, test_data');
accuracy = sum(YTest == categorical(test_labels))/numel(test_labels);
disp("测试集准确率: " + accuracy);

这里使用classify函数对测试集进行分类，然后计算准确率。

完整的代码如下：

load('dataset.mat'); % 加载数据集
num_samples = size(dataset,1);
num_train = floor(0.8*num_samples); % 80%的数据作为训练集
indices = randperm(num_samples);
train_indices = indices(1:num_train);
test_indices = indices(num_train+1:end);

train_data = dataset(train_indices, 1:end-1);
train_labels = dataset(train_indices, end);
test_data = dataset(test_indices, 1:end-1);
test_labels = dataset(test_indices, end);

layers = [
    fullyConnectedLayer(50)
    reluLayer
    fullyConnectedLayer(25)
    reluLayer
    fullyConnectedLayer(10)
    softmaxLayer
    classificationLayer
];

options = trainingOptions('adam', ...
    'MaxEpochs', 100, ...
    'MiniBatchSize', 32, ...
    'Shuffle', 'every-epoch', ...
    'Plots', 'training-progress');

net = trainNetwork(train_data', categorical(train_labels), layers, options);

YTest = classify(net, test_data');
accuracy = sum(YTest == categorical(test_labels))/numel(test_labels);
disp("测试集准确率: " + accuracy);