How to make Deep Ensemble Learning Model with Deep Learning toolbox

Can someone show me how to build an example enseble deep network like is shown in figure 2 of attached zip?

See attached or via link https://www.mdpi.com/2076-3417/10/4/1504/pdf

https://www.mdpi.com/2076-3417/10/4/1504/pdf

[Madhav Thakker]

Can you share your progress?

[Emil]

The issue was that I made the attached network I made in Deep Learning Toolbox directly referencing "Layers"

[net0,info0] = trainNetwork(inputTrain,targetTrain,lgraph_1.Layers,options);

This give errror shown below.

I should have ran it...

[net0,info0] = trainNetwork(inputTrain,targetTrain,lgraph_1,options);

This works without errors.

Kind Regards

Emil

Answer 1

[Paola Jaramillo]

Hi Emil,

Regarding the networks: RNNs, GRU and LSTMs are definitely supported.

• https://nl.mathworks.com/help/deeplearning/ug/list-of-deep-learning-layers.html

• Custom RNNs are more tricky to implement, you could use custom layers and custom training loops methods:
• https://nl.mathworks.com/help/deeplearning/deep-learning-custom-layers.html
• https://nl.mathworks.com/help/deeplearning/ref/dlarray.html

Regarding the ensemble method: the easiest way to ensemble deep learning models is to train different networks individually and then combine the predictions of all networks using one of the post-hoc ensemble approaches (e.g. average predictions or majority voting). To give you some ideas:

• https://blogs.mathworks.com/deep-learning/2019/06/03/ensemble-learning/
• https://nl.mathworks.com/help/deeplearning/ug/classify-videos-using-deep-learning.html

[Emil]

Thanks Paola

For example, the problem I have is building the layers with the deep learning toolbox.

When I try to train the network it get an error.