Why con2seq function is needed?

I am training a feedforward network. When I use con2seq function for both input and output before training, I get better network performance even it is around 2.12e^-12. However, I do not see validation or test curves in the plot. Is it normal?

Second question regarding this function, I want to predict from the trained network. Should the input values also be sequential vector? Because the network trained before in that format. Thank you in adcance.

Answer 1

[Vineet Joshi]

Hi

con2seq function helps in converting concurrent vectors to sequential vectors. The usage of both of these data structures depends on the type of problem you are solving and the network under consideration. For example: Time Series data should idealy be sequential vectors rather than concurrent.

Now that being said, in order to verify why you get better network performance with sequential vectors and why you are not seeing validation or test curves, more information on the type of network and type problem you are solving is required.

And for your second question, input values for network inference should be sequential if the input order is important.

The following documentation can help you more with the two types of data structures and when to use what.

Understanding Shallow Network Data Structures

Hope this helps.

Thanks

[Oguz Kahraman]

Thank you so much for the answer.