Rethink the tolerance for the optimization methods and the EstMethod class.

[dtnguyen2]

We need to re-think the default tolerance for the optimization methods and the EstMetho class. For lmdif, epsfcn should remain to be EPSILON (1.1920928955078125e-07) but ftol, xtol and gtol should not be EPSILON, etc... At the very least one may want to consider the tolerance within the EstMethod for the optimization method to be the same setting as eps for the EstMethod.

def lmdif(fcn, x0, xmin, xmax, ftol=EPSILON, xtol=EPSILON, gtol=EPSILON,
maxfev=None, epsfcn=EPSILON, factor=100.0, verbose=0):

def montecarlo(fcn, x0, xmin, xmax, ftol=EPSILON, maxfev=None, verbose=0,
seed=74815, population_size=None, xprob=0.9,
weighting_factor=0.8):

def neldermead( fcn, x0, xmin, xmax, ftol=EPSILON, maxfev=None,
initsimplex=0, finalsimplex=9, step=None, iquad=1,
verbose=0 ):

class EstMethod(NoNewAttributesAfterInit):

# defined pre-instantiation for pickling
config = {'sigma': 1,
          'eps': 0.01,
          'maxiters': 200,
          'soft_limits': False}

[anetasie]

@dtnguyen2 what does eps in this class do?

[dtnguyen2]

Sherpa's conf, with an appropriate coordinate transformation, is basically a 1-d root finding algorithm. So eps is the tolerance for finding said root.

[anetasie]

Is this eps only used in conf?

[dtnguyen2]

eps`` is used in confandproj```, which technically is yes :)