Skip to content
GitLab
Commit be972356 authored by Philipp Arras's avatar Philipp Arras
Browse files

Delete everything related to iteration controller logging

parent c59be760
Pipeline #61917 passed with stages
in 8 minutes and 12 seconds
Show whitespace changes
Inline Side-by-side
demos/getting_started_3.py
View file @ be972356
......@@ -103,15 +103,11 @@ if __name__ == '__main__':
data = signal_response(mock_position) + N.draw_sample()
# Minimization parameters
fo = 'energy0.txt'
fs = 'energy1.txt'
fi = 'energy2.txt'
ic_sampling = ift.AbsDeltaEnergyController(0.5, convergence_level=5,
iteration_limit=100,
file_name=fs)
ic_newton = ift.GradInfNormController(name='Newton', tol=1e-7,
iteration_limit=35, file_name=fo)
minimizer = ift.NewtonCG(ic_newton, file_name=fi)
iteration_limit=100)
ic_newton = ift.GradInfNormController(
name='Newton', tol=1e-7, iteration_limit=35)
minimizer = ift.NewtonCG(ic_newton)
# Set up likelihood and information Hamiltonian
likelihood = ift.GaussianEnergy(mean=data,
......@@ -164,5 +160,3 @@ if __name__ == '__main__':
linewidth=[1.]*len(powers) + [3., 3.])
plot.output(ny=1, nx=3, xsize=24, ysize=6, name=filename_res)
print("Saved results as '{}'.".format(filename_res))
ift.energy_history_analysis(fo, fi, fs, fname='energy_history.png')
nifty5/__init__.py
View file @ be972356
......@@ -74,7 +74,7 @@ from .minimization.energy_adapter import EnergyAdapter
from .minimization.metric_gaussian_kl import MetricGaussianKL
from .sugar import *
from .plot import Plot, energy_history_analysis
from .plot import Plot
from .library.smooth_linear_amplitude import (
SLAmplitude, LinearSLAmplitude, CepstrumOperator)
......
nifty5/minimization/descent_minimizers.py
View file @ be972356
......@@ -166,8 +166,7 @@ class NewtonCG(DescentMinimizer):
"""
def __init__(self, controller, napprox=0, line_searcher=None, name=None,
nreset=20, max_cg_iterations=200, energy_reduction_factor=0.1,
file_name=None):
nreset=20, max_cg_iterations=200, energy_reduction_factor=0.1):
if line_searcher is None:
line_searcher = LineSearch(preferred_initial_step_size=1.)
super(NewtonCG, self).__init__(controller=controller,
......@@ -175,7 +174,6 @@ class NewtonCG(DescentMinimizer):
self._napprox = napprox
self._name = name
self._nreset = nreset
self._file_name = file_name
self._max_cg_iterations = max_cg_iterations
self._alpha = energy_reduction_factor
......@@ -185,8 +183,7 @@ class NewtonCG(DescentMinimizer):
else:
ediff = self._alpha*(old_value-energy.value)
ic = AbsDeltaEnergyController(
ediff, iteration_limit=self._max_cg_iterations, name=self._name,
file_name=self._file_name)
ediff, iteration_limit=self._max_cg_iterations, name=self._name)
e = QuadraticEnergy(0*energy.position, energy.metric, energy.gradient)
p = None
if self._napprox > 1:
......
nifty5/minimization/iteration_controllers.py
View file @ be972356
......@@ -15,8 +15,6 @@
#
# NIFTy is being developed at the Max-Planck-Institut fuer Astrophysik.
from time import time
import numpy as np
from ..logger import logger
......@@ -97,15 +95,13 @@ class GradientNormController(IterationController):
"""
def __init__(self, tol_abs_gradnorm=None, tol_rel_gradnorm=None,
convergence_level=1, iteration_limit=None, name=None, p=2,
file_name=None):
convergence_level=1, iteration_limit=None, name=None, p=2):
self._tol_abs_gradnorm = tol_abs_gradnorm
self._tol_rel_gradnorm = tol_rel_gradnorm
self._convergence_level = convergence_level
self._iteration_limit = iteration_limit
self._name = name
self._p = p
self._file_name = file_name
def start(self, energy):
self._itcount = -1
......@@ -152,12 +148,6 @@ class GradientNormController(IterationController):
if self._ccount >= self._convergence_level:
return self.CONVERGED
# Write energy to file
if self._file_name is not None:
with open(self._file_name, 'a+') as f:
f.write('{} {} {}\n'.format(time(), energy.value,
self._norm(energy)))
return self.CONTINUE
......@@ -180,12 +170,11 @@ class GradInfNormController(IterationController):
"""
def __init__(self, tol, convergence_level=1, iteration_limit=None,
name=None, file_name=None):
name=None):
self._tol = tol
self._convergence_level = convergence_level
self._iteration_limit = iteration_limit
self._name = name
self._file_name = file_name
def start(self, energy):
self._itcount = -1
......@@ -218,12 +207,6 @@ class GradInfNormController(IterationController):
if self._ccount >= self._convergence_level:
return self.CONVERGED
# Write energy to file
if self._file_name is not None:
with open(self._file_name, 'a+') as f:
f.write('{} {} {}\n'.format(time(), energy.value,
crit))
return self.CONTINUE
......@@ -314,12 +297,11 @@ class AbsDeltaEnergyController(IterationController):
"""
def __init__(self, deltaE, convergence_level=1, iteration_limit=None,
name=None, file_name=None):
name=None):
self._deltaE = deltaE
self._convergence_level = convergence_level
self._iteration_limit = iteration_limit
self._name = name
self._file_name = file_name
def start(self, energy):
self._itcount = -1
......@@ -359,9 +341,4 @@ class AbsDeltaEnergyController(IterationController):
if self._ccount >= self._convergence_level:
return self.CONVERGED
# Write energy to file
if self._file_name is not None:
with open(self._file_name, 'a+') as f:
f.write('{} {} {}\n'.format(time(), energy.value, diff))
return self.CONTINUE
nifty5/plot.py
View file @ be972356
......@@ -17,7 +17,6 @@
import os
import matplotlib.pyplot as plt
import numpy as np
from . import dobj
......@@ -539,56 +538,3 @@ class Plot(object):
_plot(self._plots[i], ax, **self._kwargs[i])
fig.tight_layout()
_makeplot(kwargs.pop("name", None))
def energy_history_analysis(fname_outer, fname_inner, fname_sampling,
fname=None, figsize=[30, 10]):
"""Visualizes the behaviour of the various minimizers during a
minimization procedure.
Parameters
----------
fname_outer : string
File name of file which was written by an :class:`IterationController`
and contains the energy history of the Newton minimizer.
fname_inner : string
File name of file which was written by an :class:`IterationController`
and contains the energy history of the inverter inside of the Newton
minimizer.
fname_sampling : string
File name of file which was written by an :class:`IterationController`
and contains the energy history of the inverter which controls the
sampling.
fname : string
File name of the output plot. None lets matplotlib open an interactive
window and no file is written. Default is None.
figsize : tuple of float
figsize of output plot (see `matplotlib.pyplot.subplots`). Default is
[30, 10].
"""
# FIXME Visualize convergence criteria as well
tsa, esa, _ = np.loadtxt(fname_sampling, delimiter=' ').T
tou, eou, _ = np.loadtxt(fname_outer, delimiter=' ').T
tin, ein, _ = np.loadtxt(fname_inner, delimiter=' ').T
t0 = np.min([*tsa, *tou, *tin])
tsa = (tsa-t0)/3600
tou = (tou-t0)/3600
tin = (tin-t0)/3600
fig, (ax0, ax1) = plt.subplots(nrows=2, sharex=True, figsize=figsize)
p1 = ax0.scatter(tin, ein, marker='x', c='r', label='Newton inverter',
alpha=.3)
p2 = ax0.scatter(tsa, esa, marker='x', label='Sampling inverter', alpha=.3)
ax0.set_ylim([1.5*np.min(esa), np.max(esa) - 0.1*np.min(esa)])
ax0.set_ylabel('Conjugate gradient energy')
ax0.legend([p1, p2], [p1.get_label(), p2.get_label()])
ax1.scatter(tou, eou, marker='>', c='g')
ax1.set_ylabel('Newton energy')
# ax1.set_yscale('log')
ax1.set_xlabel("Time [h]")
plt.tight_layout()
if fname is None:
plt.show()
else:
plt.savefig(fname)
plt.close()
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment