When doing (ad hoc example)
from scipy import optimize def fit_func(x, a, b): return a*x + b optimize.curve_fit(fit_func, x_data, y_data) how can I put bounds like a>b? I know the option bounds, but it appears it does accept only explicit numbers.
Maybe some if in the definition of fit_func?
You can try with a workaround, instead of defining the function as:
def fit_func1(x, a, b): return a*x + b with constraint a>b, you can do:
def fit_func2(x, this_much_a_is_bigger_than_b, b): return (a+this_much_a_is_bigger_than_b)*x + b with constraint this_much_a_is_bigger_than_b > 0, 0 being an explicit number, and fit_func2 function is equivalent to the fit_func1 from a mathematical perspective.
Then you can have:
a = b + this_much_a_is_bigger_than_b b = b Curce_fit only supports box constraints (via least_squares). There are workarounds, but I'd rather change variables to e.g a and b-a
b --> b and a --> b + a**2