Message Boards Message Boards

GROUPS:

Classify: How does the Neural Network look like?

Posted 4 months ago
504 Views
|
2 Replies
|
4 Total Likes
|

I am working on a classification task and I want to do the work with the help of a "NeuralNetwork" classifier. There is an option to define the "NetworkDepth" by myself:

https://reference.wolfram.com/language/ref/method/NeuralNetwork.html

So the command my look like:

nn = Classify[trainSet, Method -> {"NeuralNetwork", "NetworkDepth" -> 6, ...

Later on I want to know how the Neural Networks looks like, so what kind of layers are applied. What is the sequence of the layers and so forth. How can I get these informations?

2 Replies

@Jürgen Kanz, please take a look at the guide on proper post format: http://wolfr.am/READ-1ST

You can find out about the internal NN used by simply running (in your case):

First[nn]

and then you can dig deeper with things like:

nn[[1, "Model"]]
nn[[1, "Model", "Network"]]

where you find, among many others, things like:

enter image description here

Keep in mind that the net is applied to the preprocessed data, not the data itself. To reduce preprocessing, you can use

nn = Classify[ trainSet , FeatureExtractor -> "Minimal", Method ->...]

You can also probably just design your own net:

Posted 4 months ago

Great! Thank you very much

Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
Attachments
Remove
or Discard

Group Abstract Group Abstract