Problem while converting tabel data to categorical (in order to use it for deep learning)

Hey there,

I have a 1848x4 double table of sampled data: The first three columns (named Var1 to Var3) are input data, the last one (Var4) contains the corresponding output data. There ara no NaNs.

 head(tbl):

     Var1     Var2    Var3     Var4 
    ______    ____    ____    ______

    180.09     0      100     297.72
    162.53     0      100      266.6
    144.98     0      100     233.87
    127.42     0      100     194.02
    109.86     0      100      156.1
    92.298     0      100     129.78
    74.739     0      100     107.34
     57.18     0      100     86.402

Now I want to train a multilayer perceptron (MLP) network with those data, using trainNetwork in the Deep Learning Toolbox.

To do so, I followed the implementation instructions on the Matlab help center (Train deep learning neural network - MATLAB trainNetwork - MathWorks Deutschland). My Problem now is, that I'm not able to convert the labels to categorical:

If I try to run:

labelName = 'Var4'; 
tbl = convertvars(tbl, labelName, 'categorical');
categoricalInputNames = ["Var1" "Var2" "Var3"];
tbl = convertvars(tbl, categoricalInputNames, 'categorical');

The error:

Unable to create default category names. Specify category names using the CATEGORYNAMES input argument.

occurs.

I read about this error and it seems like it happenes, if the datas values are (occasionally) quit similar to each other. Indeed, this is the case. But now I don't know how to handel this problem. Because the further steps require a categorical tabel.

I would be very pleased, If someone could help me

Answer 1

[Varun Sai Alaparthi]

Hello Philipp,

The function cannot categorize two numbers if the difference between them is less than 5e-5. You can refer to the Categorical document for more information on this.

To solve this issue, you can first apply 'discretize' function on the ‘Var4 ‘and then convert to categorical.

tbl.Var4= discretize(tbl.Var4, min(a):1e-6:max(a))

% you van even decrease the min according to your data.

labelName = 'Var4';

tbl = convertvars(tbl, labelName, 'categorical');

Run this script and this should resolve the issue.

I hope this information helps and please reach out for any further issues.

Sincerely

Varun