This pull request implements several tests for checking whether the optimization is returning valid coefficients, for both the C++ and Python versions. It aims to resolve issue #14 .
All tests start with an expression with the correct structure, but not fitted yet.
I tried to cover basic test cases where it should work properly:
sum/subtraction of two features, both with positive weights;
sum/subtraction of two features, one with negative weight;
multiplication/division of two features, with only one weight in the correct structure (to avoid having non-unique solutions and make verification easier);
a weight applied to an argument of a trigonometrical/algebraic function;
a weight applied directly to the trigonometrical/algebraic function.
For each test case, I fit the expression and perform 3 checks:
if the mean squared error is smaller than a threshold;
if all predictions are close to the correct outputs;
lacava
commented
1 year ago
This pull request implements several tests for checking whether the optimization is returning valid coefficients, for both the C++ and Python versions. It aims to resolve issue #14 .
All tests start with an expression with the correct structure, but not fitted yet.
I tried to cover basic test cases where it should work properly:
For each test case, I fit the expression and perform 3 checks: