combining classification from 2 different NN , on two different datasets

we have two diffrent datasets to be utilised for 2 different NN (classification type) and decision be made on results of both,eg if one has got defect in eye (retina dataset and iris dataset)

how to implement in the code?

Answer 1

[Abhishek Gupta]

Hi,

As per my understanding, you want to train two different Neural Network on two different datasets for the classification problem. This task can be achieved by implementing and training two neural networks separately on the two datasets. After training, you can use these trained networks to make the prediction and then can compare the predictions at the same time.

The following documentation would help you in creating a simple network for classification: -

Create Simple Deep Learning Network for Classfication

[Yogini Prabhu]

thanks for reply!

well its not about comparing, I need to make a single decisions out of the two classifications

its like looking at two separate aspects of a perticular issue and collaborating them towards one decisison.

[Abhishek Gupta]

I see. One way to get a single prediction would be as follows: -

1. Create a network for each dataset to convert your input into a feature vector (let's say of size 10x1 for each dataset).
2. Combine the feature vectors from the above two networks into one vector (concatenate the vectors to get 20x1 column vector). This combined vector will contain the input features from the two datasets.
3. Use the combined feature vector as your input and train another network to make your final predictions, whose output would be a single value.

I hope this would help!

[Yogini Prabhu]

Hi,

your points in the reply are indeed going to solve the requirement.

the thing is, I have used nprtool (Neural pattern recognition app) of Matlab and have obtained following results:-

so I request you to help me find the feature vector, out of all the components within the mat file obtained.

[Abhishek Gupta]

For your problem, I would suggest you use Deep Network Designer instead, which will help you to create a customized network as described above. Here is the documentation link which might help you in getting started: -

https://www.mathworks.com/help/deeplearning/gs/get-started-with-deep-network-designer.html

[Yogini Prabhu]

thanks for suggesting the way to Deep network

but with a toddler baby, its difficult to migrate to a new area, (and I wish to achieve deadline of research paper to a quality submission ). i have used the app of nprtool (as said above).

kindly help with the results got, in the attached mat file