Matlab trainNetwork的问题。设置观测值格式的正确方法是什么？

% 生成示例数据
numObservations = 100;
numFeatures = 5;
data = rand(numObservations, numFeatures);

% 创建网络
layers = [
    imageInputLayer([numFeatures 1]) % 输入层，形状为[numFeatures 1]
    fullyConnectedLayer(10) % 全连接层
    reluLayer % ReLU激活层
    fullyConnectedLayer(1) % 输出层
];

% 训练网络
net = trainNetwork(data, data, layers);

% 生成示例数据
numObservations = 50;
numTimeSteps = 10;
numFeatures = 3;
data = rand(numObservations, numTimeSteps, numFeatures);

% 创建网络
layers = [
    sequenceInputLayer(numFeatures) % 输入层，形状为[numFeatures]
    lstmLayer(10, 'OutputMode', 'last') % LSTM层
    fullyConnectedLayer(1) % 输出层
];

% 训练网络
net = trainNetwork(data, data, layers);

% 生成示例数据
numObservations = 20;
imageSize = [64 64];
numChannels = 3;
data = rand(numObservations, imageSize(1), imageSize(2), numChannels);

% 创建网络
layers = [
    imageInputLayer(imageSize) % 输入层，形状为[imageSize numChannels]
    convolution2dLayer(3, 8) % 卷积层
    reluLayer % ReLU激活层
    maxPooling2dLayer(2, 'Stride', 2) % 最大池化层
    fullyConnectedLayer(1) % 输出层
];

% 训练网络
net = trainNetwork(data, data, layers);