Error using trainNetwork, Invalid training data. Predictors must be a N-by-1 cell array of sequences, where N is the number of sequences. All sequences must have the same feat

Question

Goh Dione 2022-2-9

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1646405-error-using-trainnetwork-invalid-training-data-predictors-must-be-a-n-by-1-cell-array-of-sequences

回答： Milan Bansal 2023-10-3

I am trying to train a deep learning model with the training data and training label both having the same size of 409600*1. However, i keep getting the following error:

Error using trainNetwork (line 184)

Invalid training data. Predictors must be a N-by-1 cell array of sequences, where N is the number

of sequences. All sequences must have the same feature dimension and at least one time step.

Error in ML (line 71)

trainedNet = trainNetwork(y_ML(:,1)',Label_train',ly,options);

Am I doing something wrong?

clear all;close all; clc;

%% Load datasets

% load('MLsample.mat');

%Original signal

load('x_ML.mat') %label with size (409600,3)

n_samples = size(x_ML,1); %number of training samples

%received signal

load('y_ML.mat') %data with size (409600,3)

% example = matfile('x1_ML.mat')

% C = example.x1_ML;

%power allocation factor = 0.8

%% Partition Data

%Split the data into training, validation, and test sets.

%First, reserve approximately one third of the observations for the test set. Then, split the remaining data in half to create the training and validation sets.

Ds = [x_ML(:,1), y_ML(:,1)]; %Label, Data set

vector = randperm(n_samples);

% Random Split the Dataset into test and validation set

Ds_Vad = zeros(30,size(Ds,2));

Ds_Test = zeros(30,size(Ds,2));

for i = 1:60

if i <= 30

Ds_Test(i,:)= Ds(vector(i),:);

else

Ds_Vad(i-30,:) = Ds(vector(i-30),:);

end

Xdata_Vad = Ds_Vad(:,(2)); %data = column2 onwards

Ylabel_Vad = Ds_Vad(:,(1)); %label = column1

Xdata_Test = Ds_Test(:,(2)); %data = column2 onwards

Ylabel_Test = Ds_Test(:,(1)); %label = column1

Label_train = categorical(x_ML(:,1));

Ylabel_Vad = categorical(Ylabel_Vad);

Ylabel_Test = categorical(Ylabel_Test);

%% Network Setting

ly = [ ...

sequenceInputLayer(33)

tanhLayer

fullyConnectedLayer(10)

tanhLayer

fullyConnectedLayer(2)

softmaxLayer

classificationLayer];

maxEpochs = 2000;

miniBatchSize = 27;

%% Learning Options Setup for Fully Connected Network

options = trainingOptions('adam', ...

'MaxEpochs',maxEpochs, ...

'LearnRateDropFactor',0.01, ...

'ValidationData',{Xdata_Vad',Ylabel_Vad'}, ...

'ValidationFrequency',20, ...

'MiniBatchSize',miniBatchSize, ...

'Verbose',true, ...

'Plots','training-progress');

%% Training FC_net

trainedNet = trainNetwork(y_ML(:,1)',Label_train',ly,options);

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

yanqi liu 2022-2-10

yes，sir，may be upload your data to debug

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Milan Bansal 2023-10-3

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1646405-error-using-trainnetwork-invalid-training-data-predictors-must-be-a-n-by-1-cell-array-of-sequences#answer_1323910

在 MATLAB Online 中打开

Hi,

As per my understanding you are facing an error while training a Neural Network using "trainNetowork" function while performing feature-label classification.

For feature-label classification, the input layer in the network must be a "featureInputLayer" instead of "sequenceInputLayer". Modify the input layer in the network as shown in the code below.

ly = [ ... 
    featureInputLayer(1) % number of features in training data 
    tanhLayer 
    fullyConnectedLayer(10) 
    tanhLayer 
    fullyConnectedLayer(2) % number of classes 
    softmaxLayer 
    classificationLayer]; 