Optimization with few data points

100 Views Asked by At

I am facing some trouble working on a problem. I am trying to optimize a function with 34 data points (no possibility to acquire new data points in the near future) whose input is 39 dimensional and output is a single dimension scalar.

I was thinking of applying Bayesian Optimization for this but since I cannot evaluate the function at new points, is it still feasible?

What are some other models I can apply for this problem?

What if I just want to see the if there is a correlation between the input and output? (Can't use pearson correlation because it's just between a pair of variables, here I have 39+1 variables)

Hope I can get help here. Thanks in advance!