Train_X must be a float.

May anybody help me to train my autoencoder for training dataset.

My training dataset is of size 10000*3072.But I want to let my model to train for its subset of different sizes like 1000*700, 1500*1000, 784,100 etc

but for every dataset sometimes it generates error: train_x must be float or train_x must be an integer.

here is the error.

here is the code

function demo_SAE
load('data_batch_1.mat');

train_x=(data(1:1200,1:1200));
rand('state',0)
sae = saesetup([784 100]);
sae.ae{1}.activation_function       = 'sigm';
sae.ae{1}.learningRate              = 1;
sae.ae{1}.inputZeroMaskedFraction   = 0.0;
opts.numepochs =   10;
opts.batchsize = 100;
sae = saetrain(sae, train_x, opts);
visualize(sae.ae{1}.W{1}(:,2:end)')

[Walter Roberson]

What is class(train_x) ?

[Rizwan Hassan]

train_x=training data set

[Walter Roberson]

Use the GUI to put a breakpoint at the beginning of the line

rand('state',0)

Now execute the code. It will stop before executing the rand() call. Now at the command window, give the command

class(train_x)

and copy and paste the result into a response here.

Answer 1

[Puru Kathuria]

Hi,

I see that you are not able to train your model because of the data type of your training data. I think the following snippet might solve your issue. Put the following snippet before training and see if it helps.

isItFloat = isfloat(train_x);
if(isItFloat == 0)
    train_x = double(train_x);
end