symbolic_regression_part1_bis

Symbolic regression - Multiple variables

Extension to multiple variables is straight-forward.

For example consider the $f(x,y) = ln(x^2 + y^2)$ function:

We only need to add a column to the input data:

std::istringstream training(R"(
  -2.079, 0.25, 0.25
  -0.693, 0.50, 0.50
   0.693, 1.00, 1.00
   0.000, 0.00, 1.00
   0.000, 1.00, 0.00
   1.609, 1.00, 2.00
   1.609, 2.00, 1.00
   2.079, 2.00, 2.00
)");

and a function to the function set:

prob.insert<vita::real::ln>();

(for your ease the above code is in the examples/symbolic_regression01.cc file)

and what we get is:

[INFO] Reading dataset from input stream...
[INFO] ...dataset read. Examples: 8, categories: 0, features: 2, classes: 0
[INFO] Setting up terminals...
[INFO] ...terminals ready. Variables: `X1` `X2`
[INFO] Number of layers set to 1
[INFO] Population size set to 100
Run 0.     0 (  0%): fitness (-107.645)
Run 0.     0 (  1%): fitness (-102.444)
Run 0.     0 (  3%): fitness (-88.5616)
Run 0.     0 ( 16%): fitness (-88.4484)
Run 0.     0 ( 19%): fitness (-67.8505)
Run 0.     0 ( 27%): fitness (-49.6343)
Run 0.     1 ( 87%): fitness (-31.0667)
Run 0.    20 ( 42%): fitness (-23.9684)
Run 0.    48 (  6%): fitness (-20.1796)
Run 0.    51 ( 28%): fitness (-0.0174211)
[INFO] Elapsed time: 0.068s
[INFO] Training fitness: (-0.0174211)

CANDIDATE SOLUTION
log(((X1*X1)+(X2*X2)))

FITNESS
(-0.0174211)

---

GO ON TO PART 2 →