How to use GA (Genetic Algorithm) without explicit objective function and with data-set?

Dear scholars,

Suppose I have experimental data-set. I have also the quantitative values of the inputs and output. But I donn't have the objective function. My questions is, does GA still work? I means how to implement genetic algorithm with only data-set but not the objective function?

----

Can I generate neural netwok model for new data prediction and then export the model to integrate it with the GA?

Thanks!

[Stephen23]

Why bother? If you have the data set, just use min to find the minimum.

Or are you hoping to perform some kind of non-linear interpolation (which is essentially inventing new data) which could conceivably return a global minimum at some other location than that of the minimum data value?

[Alireza Babaei]

I need to go through optimization techniques since my experimental data-set is limitted to some specific inputs only.

[Walter Roberson]

Okay, so you have some input data. Possibly it might even consist of inputs and responses. What do you need to minimize?

[John D'Errico]

You cannot. I think you do not understand optimization. If all you have are a list of data points, then you do not have a function you can optimize. You just have a list of data points, nothing more.

An optimization tool requires an objective function that it can minimize (or maximize), providing a function value at ANY intermediate point.

Stephen already told you what you can do: that is, compute the minimum value over the set of points you have.

You have then said you need to go through optimization techniques. But I'm sorry, that makes no sense here.

Finally, Stephen also suggested you might try to interpolate the data. That is fraught with its own issues, because you will need to understand the interpolation methods used before that will help you any.

Finally, you might also be interested in response surface optimization. But again, I think you will need to do some reading and thinking about what you have before that will help you.

[Walter Roberson]

My suspicion is that the poster needs something equivalent to Neural Network approaches for either classification or forecasting, but was hoping to find some non-NN approach in order to cross-check (or just because they have to study a number of different aproaches to reach the same end.)

Answer 1

[Pratyush Roy]

Hi,

Genetic algorithm will not work without an optimization function. To understand how GA works, refer to this link.

Hope this helps!