Simple numerical networks

I’ve had a go at trying to use Deep Cognition for some basic numerical data but am struggling with finding tutorials or information that will help me decide on what error functions and learning programs are appropriate.

Currently I can get simple multi-layer networks to train, but there is no accuracy data present (the chart stays blank, when flipping to other screen then back to the chart the error rate is stuck at 0).

I am also unable to model multiple outputs, I cant seem to find the place where I tell DeepCognition how many outputs there needs to be, so i’m unsure how to properly set up classification of numerical inputs.

What input and output ranges are supported, how can I get accuracy feedback when training and are there any good tutorial resources on how to use DeepCognition properly? (most i’ve found so far are quick guides, parrot learning that doesn’t really explain much about the functions, what they do and how to use them properly)

DLS supports 1000 categorical input and Single output.
Your output can be a class array or numeric whatever you want but It supports only 1 output.
Yes there is great course on udemy available which will guide you on how to use different functions with DLS.

and by Multiple output if you mean multiple class for classification problem then you can have upto 1000 class and use categorical datatype.
If you could share some details of your input and output then it would be great.
Here you can find articles on problem solving using DLS


I noticed in your example, the accuracy chart also does not seem to work. Is that normal? Can it be fixed?

In which example Stock prediction?
That is a regression problem that’s why there is no accuracy graph.

1 Like