an aborted fit with `method='least_squares'` does not report correct fit results and statistics

[newville]

Description

As shown in #894, a fit with method='least_squares' that aborts, perhaps due to too many function evaluations, does not properly evaluate the residual and fit statistics with "last/best available parameter values".

Note that other methods do evaluate the residual and fit statistics with "last/best available parameter values" when the fit is aborted.

A Minimal, Complete, and Verifiable example

see #894

Link(s)

see #894

[eendebakpt]

@newville Thanks for picking this up. It seems also basinhopping, ampgo and shgo have the same issue. Tested with:

import lmfit
from lmfit import Model
import numpy as np

def gaussian(
    x, mean: float, std: float, amplitude: float = 1.0, offset: float = 0.0):
    x0 = x - mean
    y = offset + amplitude * np.exp(-x0 * x0 / (2.0 * std * std))
    return y


x_data, y_data = (np.array([18.025, 18.075, 18.125, 18.175, 18.225, 18.275, 18.325, 18.375,
        18.425, 18.475, 18.525, 18.575, 18.625, 18.675, 18.725, 18.775,
        18.825, 18.875, 18.925, 18.975, 19.025, 19.075, 19.125, 19.175,
        19.225, 19.275, 19.325, 19.375, 19.425, 19.475, 19.525, 19.575,
        19.625, 19.675, 19.725, 19.775, 19.825, 19.875, 19.925, 19.975,
        20.025, 20.075, 20.125, 20.175, 20.225, 20.275, 20.325, 20.375,
        20.425, 20.475, 20.525, 20.575, 20.625, 20.675, 20.725, 20.775,
        20.825, 20.875, 20.925, 20.975]), np.array([ 43,  27,  14,  16,  20,  10,  15,   8,   3,   9,   2,   4,   8,
          3,   3,   1,   2,   4,   5,   0,   1,   3,   1,   0,   0,   1,
          0,   0,   0,   0,   0,   0,   1,  23, 202, 824, 344, 279, 276,
         23,   2,   7,   8,  17,  22,   3,  25,  24,  17,  99, 288, 172,
        252, 103,  24,  20,   4,   1,   3,   1], dtype=int))

initial_parameters = {'mean': 18.286083743842365, 'std': 20.07653394255875, 'amplitude': 43}

m=Model(gaussian)
m.set_param_hint("amplitude", min=0)

# not tested: 'differential_evolution', 'brute','newton', 'trust-ncg','trust-krylov','dogleg','dual_annealing' 
for method in ['leastsq','least_squares','basinhopping','ampgo','nelder','lbfgsb','powell','cg','cobyla'
              ,'bfgs','shgo',    ]:
    r=m.fit(y_data, **initial_parameters, x=x_data,  method=method, max_nfev=2)
    #print(r.success,r.message)
    print(method, r.redchi) # has no valid value

[newville]

@eendebakpt Thanks for catching that I guess that should not be too surprising that other methods have a similar problem.

[newville]

@eendebakpt I adapted the PR at #897 to handle (and test) aborted fits for more methods, and make sure that they also end up with clearly aborted but still not-useless values for fit statistics.