Where can I find the datasets "digitTrai​nCellArray​Data"

Hi all,

Now I am running the MATLAB sample code for "Train Stacked Autoencoders for Image Classification". The codes are as below:

% Load the training data into memory
[xTrainImages,tTrain] = digitTrainCellArrayData;

% Display some of the training images
clf
for i = 1:20
    subplot(4,5,i);
    imshow(xTrainImages{i});
end

However, I can't find the dataset. Here is the error:

Undefined function or variable 'digitTrainCellArrayData'.

Can anybody share the dataset? Thanks!

PS. I am using MATLAB 2015b, 64bits, and my machines are: (1)MacBook Pro;(2)Precision T5810(CPU E5-1620 v3 @ 3.50GHz,RAM 32G, Win10 64 bits)

[Harish Iyer]

Hi,

The the below line, loads input data into two variables xTrainImages and tTrain.

[xTrainImages,tTrain] = digitTrainCellArrayData;

xTrainImages is quite clear that it is an array of 5000 input images size - 28x28 each.

However, I am not able to undertand what is tTrain. It is a 10x5000 2D array which most certainly looks like labels. Looking at places it is used later in the code. However it is not clear why 5000 0's and 1's are used as a label? How do we identify which row represents which label?

The reason I want to understand is that I want to experiment with other image datasets which will have to be converted in this format.

Is there any explaination what is contained in this variable tTrain?

Thanks,

Harish

[Walter Roberson]

labels are associated with columns not rows. Rows are features, columns are samples.

Answer 1

[Walter Roberson]

That example and dataset are from R2016b and do not exist in earlier versions.

[Chaya N]

I'm running R2015b on a Mac. This example is most certainly available and works just fine.

[Walter Roberson]

The R2015b example uses digittrain_dataset instead of digitTrainCellArrayData as can be seen at http://www.mathworks.com/help/releases/R2015b/nnet/examples/training-a-deep-neural-network-for-digit-classification.html

[Chaya N]

Aah, you're right! My bad!

[yiming du]

I am using 2016a, is there a way to get the dataset and get the example running? thnaks!

[Walter Roberson]

Yes, you can upgrade to R2016b to get it

[Jianshe Feng]

Thanks guys~ I am using the 2016b now.

[zina tayati]

how could i upgrade in this way

[SUMOD SUNDAR]

i am using 2018a.

In which folder I can locate digitTrainCellArrayData.

[Walter Roberson]

digitTrainCellArrayData turns out to be a .m file that goes through a couple of steps to load a .csv into a table.

In R2018b (I happened to have that one handy, but did not have R2018a handy),

fullfile(matlabroot, 'toolbox', 'nnet', 'nndemos', 'nndatasets', 'digitTrainCellArrayData.m')

and the file it loads (and transforms) is at

fullfile(matlabroot, 'toolbox', 'nnet', 'nndemos', 'nndatasets', 'DigitDataset', 'digitTrain.csv')

Answer 2

[Nidhi Chakravarthy]

This is the mnist database(Modified National Institute of Standards and Technology database).

Here is the link to it: https://www.bing.com/search?q=mnist%20database&FORM=AFTC02&PC=AFTC