Error when training neural network with multi feature inputs

Hi,

I am currently working on a project in which I use a neural network with 3 feature inputs. The first one (localPosScore) is a 1x9 vectors and the others (msgX) are 1x7 vectors :

I create an arrayDatastore from a cell array with my training datas and the read function gives me this result :

However when I use the trainNetwork function the command answers me this error

Error using trainNetwork (line 184)

Input datastore returned more than one observation per row for network input 1.

I don't where I miss something because when I read the trainNetwork's help page I have a cell array of {numInputs+1} elements with a categorical label.

May you help me to solve this probem ?

Thanks in advance

Mathieu

[chechezhe Chen]

https://www.mathworks.com/support/search.html/answers/1587899-how-to-create-a-datastore-for-using-the-deep-network-designer-app.html?fq[]=asset_type_name:answer&fq[]=category:deeplearning/deep-learning-with-time-series-sequences-and-text&page=1

[YOUNGWOO AN]

Did you solve this problem?

[Muhammad]

This is how i fixed this probelm:

this was my before datastore

>> preview(combined_ds)

ans =

1×4 cell array

{32×32 uint8} {32×32 uint8} {1×80 double} {1×1 cell}

you can see that the linear layer has size 1x80.. this is causing the problem.. try to make it 80x1.. this will solve the problem..

Answer 1

[Prince Kumar]

Hi,

Please refer to the following answers:

• https://www.mathworks.com/matlabcentral/answers/522086-how-can-i-train-multi-input-deep-network-without-datastore
• https://www.mathworks.com/matlabcentral/answers/489864-how-to-train-deep-learning-network-with-multi-inputs

Hope this helps!!