Neural Network design problem

Hi

This will be my first ever neural network code. I am trying to implement feedforward netwrok. Let me explain what I am trying to achive.

I have a 1D vector P. This is the input to my network training. In the first round of network estimates(I have not decided yet how many layers will there be), a new vector P1 come as an output. Now i need to plug in the output to formula, say, Q=P1*scalar. I already have predefined Q range, say between 10 to 20. if the output P1, gives Q within this range, the network training needs to be stopped, else it should go for second round.

I have the following questions

1. How to make this work
2. On m furst round once the P1 does not meet the Q range, do i need to feed the updated P1, for second round of training

I appreciate the help in advance

Answer 1

[Vineet Joshi]

Hi

It looks like you are trying to scale your value P to a value Q where Q is between 10-20 and Q = P1 * scaler.

If you want to make this work, to being with, you need some labelled data to train the network.

The data can have the following columns.

P ------- Expected Q ------- Expected P1 (= Expected Q / scaler).

Once you have this data, you can fit a neural network between P (Inputs) and Expected P1 (Targets).

You can refer the following documentation to learn more about that.

Fit Data with a Shallow Neural Network

You may also refer the following resources for more information.

• How Do Neural Networks Work?
• Workflow for Neural Network Design

Hope this helps.