Artificial Neural Network - 2 inputs, 11 outputs

Hi,

I have 2 numerical inputs, Time period and Ductility demand, and 11 outputs consisting of ratios of total weight for 10% increments in height and 1 total weight. Training data for different combinations of Time period and ductiltiy demand and their resulting ratios and total weight have been collected.

My task is to create a Neural Network where a designer can input the Time period and Ductility demand which the network gives the correct height ratios and total weight.

What is the best method to use to complete my task? I have experimented with the deep network designer but I have not made progress.

Thanks,

Alex

Answer 1

[Shashank Gupta]

Hi Alex,

You can tackle this problem in many different way, all you need to do is form a proper deep neural network and train the model with the data available to you. There are resouces available to do the same. Check out this link which talk about time series data and regression, You may also wish to look at this link to create a simple regression model for your problem. Its possible that the data which you have are complex in nature and could not be easily generalised by typical neural network, in such scenerios you have to experiment with increasing number of layer or using more time dependence model such as RNN or LSTM. Before you do all these experiment make sure to visualise the input and output data properly. This will help you decide on what sort of experiments are needed. Don't blindly do experiment. Its also possible that you may not even need to form a deep network instead the job can be done by simply curve fitting or SVR.

I hope this helps you. Keep exploring. Do reach out if you stuck anywhere.

Cheers.