Calling a Function specific from a ToolBox

Hi again Mathworks people. :)

I am attempting to use the function classify() from the Deep Learning Toolbox, however it seems to call instead on the Statistic and Machine Learning function with the same name.

My question is: is there a way to specify where the function/which toolbox the function is called from?

this is the function I wish to use:

https://au.mathworks.com/help/deeplearning/ref/seriesnetwork.classify.html

this is the function it is currently calling (which I do not want to use):

https://au.mathworks.com/help/stats/classify.html?searchHighlight=classify&s_tid=srchtitle_classify_2

Thanks in advance!

Miller

Answer 1

[Steven Lord]

The function you indicated you wanted to call is only called in the event that the first input argument is either "a SeriesNetwork or a DAGNetwork object". Is it? Let's see which version of the train function you're calling.

compnet             = competlayer(3);
which -all train(compnet, rand(3))

/MATLAB/toolbox/nnet/nnet/@network/train.m  % network method

This function returns a "Trained network, returned as a network object." A network object is neither a SeriesNetwork nor a DAGNetwork. What can you do with a network object?

methods network

Methods for class network:

adapt         disp          genFunction   init          loadobj       perform       revert        stack         subsref       unconfigure   
configure     display       gensim        isconfigured  network       prune         sim           subsasgn      train         view          

What happens when you try to classify a network?

which -all classify(network)

/MATLAB/toolbox/stats/stats/classify.m

If you want to use the classify function you need to work with a deep neural network, not a shallow neural network.

[MILLER BIGGELAAR]

Can you please provide some helpful links so I can begin creating an operational one?

[Steven Lord]

I don't know exactly what you're trying to do, so the links to the SeriesNetwork and DAGNetwork pages in the first paragraph of my answer would probably be my first stop.

Alternately look for the sub-category on the list of functions in Deep Learning Toolbox to find the one that closest describes the application you're working on (processing image data, processing time series data, approximating a function, etc.) and see if there are pretrained networks or network creation functions in that sub-category.

Answer 2

[Chunru]

seriesnetwork is a calss. It has a method of "classify". In order to use the method, you need to create the network and train it. Then can call classify by

classify(net, ...)   % where net is the network trained

[MILLER BIGGELAAR]

Will you look at my code?

[Chunru]

You post your code to see who can help you.

Answer 3

[MILLER BIGGELAAR]

I have already, here is my code:

clear
clc

test                = Iris_data;
label               = zeros(150,1);
label(1:50,:)       = 1;
label(51:100,:)     = 2;
label(101:150,:)    = 3;
test(:,5)           = label;
k                   = randperm(150,50);
trainset            = test(k(1:50),:);
trainsetlabel       = test(k,5);
test(k,:)           = [];
datalabel           = test(:,5);
%%%%
compnet             = competlayer(3);
compnet             = train(compnet,trainset');
outputs             = compnet(trainset');
classes             = vec2ind(outputs);
testingnet          = classify(compnet,test')

Attached is the data