MATLAB Answers

Error using trainNetwork - invalid training data

5 views (last 30 days)
Jinyang Du
Jinyang Du on 29 Jul 2021
Commented: KSSV on 29 Jul 2021
layers = [
sequenceInputLayer(27)
bilstmLayer(100,"OutputMode","last")
fullyConnectedLayer(3);
softmaxLayer();
classificationLayer()
];
options = trainingOptions("adam", ...
"MaxEpochs",250, ...
"InitialLearnRate",0.005, ...
"GradientThreshold",1, ...
"Shuffle","every-epoch", ...
"Plots","training-progress", ...
"LearnRateDropPeriod",200,...
"LearnRateSchedule","piecewise");
The network I used is written above. And I tried to use net = trainNetwork(XTrain,YTrain,layers,options); to train this network. But it seems never work (I have converted table to cell before training). The training data is attached. The XTrain data is a 33750 x 27 predictor and YTrain is a 33750 x 1 respond. I intend to use this network to classify these data (predictors in X map into categories in Y). Could anyone show me how to get going? I am quite deserate by now.
  2 Comments
Jinyang Du
Jinyang Du on 29 Jul 2021
I just used
x = readtable('XTrain.csv');
y = readtable('YTrain.csv');
XTrain = table2cell(x);
YTrain = table2cell(y);
net = trainNetwork(XTrain,YTrain,layers,options);

Sign in to comment.

Accepted Answer

KSSV
KSSV on 29 Jul 2021
x = readtable('XTrain.csv');
y = readtable('YTrain.csv');
XTrain = table2array(x);
YTrain = table2array(y);
XTrain = num2cell(XTrain,2) ;
XTrain = cellfun(@transpose,XTrain,'UniformOutput',false) ;
YTrain = categorical(YTrain) ;
layers = [
sequenceInputLayer(27)
bilstmLayer(100,"OutputMode","last")
fullyConnectedLayer(3);
softmaxLayer();
classificationLayer()
];
options = trainingOptions("adam", ...
"MaxEpochs",250, ...
"InitialLearnRate",0.005, ...
"GradientThreshold",1, ...
"Shuffle","every-epoch", ...
"Plots","training-progress", ...
"LearnRateDropPeriod",200,...
"LearnRateSchedule","piecewise");
net = trainNetwork(XTrain,YTrain,layers,options);

More Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!