Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Classes for making standard monitor functions #114

Merged
merged 22 commits into from
Jan 14, 2025
+602 −67
Merged

Classes for making standard monitor functions #114

Show file tree
Hide file tree
Changes from 16 commits
Commits
Show all changes
22 commits
Select commit Hold shift + click to select a range
770292f
MonitorFactory and some basic subclasses
jwallwork23 Aug 25, 2024
2ba8628
Add monitor module to star imports
jwallwork23 Aug 25, 2024
ee54b3f
Unit test for ConstantMonitorFactory
jwallwork23 Aug 25, 2024
d44fbef
Unit tests and TODOs
jwallwork23 Aug 25, 2024
c9d6caf
Docstrings; gradient monitor
jwallwork23 Aug 25, 2024
2187d07
Solution-based monitors
jwallwork23 Aug 25, 2024
f66e7c0
More rigorous testing
jwallwork23 Aug 25, 2024
4f12b6b
GradientHessianMonitorFactory
jwallwork23 Aug 26, 2024
49a4d15
Formatting
jwallwork23 Aug 26, 2024
0da2be1
Combined version
jwallwork23 Aug 26, 2024
c393aa5
Factory->Builder
jwallwork23 Aug 26, 2024
3aa4e3d
Project solution field onto current mesh before recovery
jwallwork23 Aug 26, 2024
2aacae8
Lint
jwallwork23 Dec 24, 2024
1432347
Use BallMonitorBuilder in monge_ampere1 demo
jwallwork23 Dec 24, 2024
5c27f6c
Implement and use RingMonitorBuilder
jwallwork23 Dec 28, 2024
dfb651f
Rename mf->mb; check __call__ method; no cover for abstractmethod
jwallwork23 Dec 28, 2024
5d7d116
Apply suggestions from @ddundo code review
jwallwork23 Jan 13, 2025
8e94b1e
Make monitor method private; inner needs ufl
jwallwork23 Jan 13, 2025
9cda85f
Use radius 0.4 in ma demo
jwallwork23 Jan 13, 2025
62dfbce
Merge branch 'main' into 96_monitor-utils
jwallwork23 Jan 13, 2025
dc0a7b1
Reorder init for ball monitor class
jwallwork23 Jan 14, 2025
a6636e0
Update solver progress after changing ring radius
jwallwork23 Jan 14, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 10 additions & 13 deletions demos/monge_ampere1.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,6 +41,7 @@
#
# We begin the example by importing from the namespaces of Firedrake and Movement. ::

import numpy as np
from firedrake import *

from movement import *
Expand Down Expand Up @@ -73,21 +74,17 @@
#
# for some values of the parameters :math:`\alpha`, :math:`\beta`, and :math:`\gamma`.
# Unity is added at the start to ensure that the monitor function doesn't get too
# close to zero.
# close to zero. Here we can think of :math:`\alpha` as relating to the amplitude of the
# monitor function, :math:`\beta` as relating to the width of the ring, and
# :math:`\gamma` as the radius of the ring.
#
# Here we can think of :math:`\alpha` as relating to the amplitude of the monitor
# function, :math:`\beta` as relating to the width of the ring, and :math:`\gamma` as
# the radius of the ring.


def ring_monitor(mesh):
alpha = Constant(20.0)
beta = Constant(200.0)
gamma = Constant(0.15)
x, y = SpatialCoordinate(mesh)
r = (x - 0.5) ** 2 + (y - 0.5) ** 2
return Constant(1.0) + alpha / cosh(beta * (r - gamma)) ** 2
# For convenience, Movement provides a builder class for ring monitors, amongst other
# commonly used monitor functions. ::
jwallwork23 marked this conversation as resolved.
Show resolved Hide resolved

mb = RingMonitorBuilder(
centre=(0.5, 0.5), radius=np.sqrt(0.15), amplitude=20.0, width=200.0
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have a feeling that using radius=np.sqrt(0.15) might be confusing and lead people to think they need to input the sqrt of the radius. Why not just input e.g. radius=0.4?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Addressed in 9cda85f.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks! But the optimisation callback should be updated too. It looks like this now:

   0   Volume ratio 12.91   Variation (σ/μ) 1.21e+00   Residual 1.16e+00
   1   Volume ratio  6.77   Variation (σ/μ) 6.15e-01   Residual 4.85e-01
   2   Volume ratio  5.84   Variation (σ/μ) 5.69e-01   Residual 4.26e-01
   3   Volume ratio  6.34   Variation (σ/μ) 5.14e-01   Residual 3.74e-01
   4   Volume ratio  6.38   Variation (σ/μ) 4.88e-01   Residual 3.22e-01
   5   Volume ratio 12.05   Variation (σ/μ) 4.55e-01   Residual 2.57e-01
   6   Volume ratio 11.69   Variation (σ/μ) 4.26e-01   Residual 2.30e-01
   7   Volume ratio 11.98   Variation (σ/μ) 4.23e-01   Residual 1.98e-01
   8   Volume ratio 11.81   Variation (σ/μ) 4.17e-01   Residual 1.95e-01
   9   Volume ratio 12.00   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  10   Volume ratio 12.07   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  11   Volume ratio 12.09   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  12   Volume ratio 12.09   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  13   Volume ratio 12.09   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  14   Volume ratio 12.09   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  15   Volume ratio 12.09   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  16   Volume ratio 12.09   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  17   Volume ratio 12.09   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  18   Volume ratio 12.09   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  19   Volume ratio 12.09   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  20   Volume ratio 12.09   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  21   Volume ratio 12.09   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  22   Volume ratio 12.09   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  23   Volume ratio 12.09   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  24   Volume ratio 12.09   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  25   Volume ratio 12.09   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  26   Volume ratio 12.10   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  27   Volume ratio 12.18   Variation (σ/μ) 4.16e-01   Residual 1.94e-01
  28   Volume ratio 12.38   Variation (σ/μ) 4.16e-01   Residual 1.93e-01
  29   Volume ratio 12.43   Variation (σ/μ) 4.16e-01   Residual 1.93e-01
  30   Volume ratio 12.44   Variation (σ/μ) 4.16e-01   Residual 1.93e-01
  31   Volume ratio 12.44   Variation (σ/μ) 4.16e-01   Residual 1.93e-01
  32   Volume ratio 12.44   Variation (σ/μ) 4.16e-01   Residual 1.93e-01
  33   Volume ratio 12.44   Variation (σ/μ) 4.16e-01   Residual 1.93e-01
  34   Volume ratio 12.44   Variation (σ/μ) 4.16e-01   Residual 1.93e-01
  35   Volume ratio 12.44   Variation (σ/μ) 4.16e-01   Residual 1.93e-01
  36   Volume ratio 12.44   Variation (σ/μ) 4.16e-01   Residual 1.93e-01
  37   Volume ratio 13.34   Variation (σ/μ) 4.18e-01   Residual 1.94e-01
  38   Volume ratio 15.21   Variation (σ/μ) 4.38e-01   Residual 2.09e-01
  39   Volume ratio 17.20   Variation (σ/μ) 4.76e-01   Residual 2.22e-01
  40   Volume ratio 11.89   Variation (σ/μ) 4.30e-01   Residual 1.45e-01
  41   Volume ratio 12.41   Variation (σ/μ) 4.13e-01   Residual 1.01e-01
  42   Volume ratio  8.76   Variation (σ/μ) 3.79e-01   Residual 7.36e-02
  43   Volume ratio 10.00   Variation (σ/μ) 3.85e-01   Residual 6.36e-02
  44   Volume ratio  8.57   Variation (σ/μ) 3.65e-01   Residual 5.17e-02
  45   Volume ratio  9.48   Variation (σ/μ) 3.72e-01   Residual 4.17e-02
  46   Volume ratio  8.70   Variation (σ/μ) 3.60e-01   Residual 3.51e-02
  47   Volume ratio  9.31   Variation (σ/μ) 3.64e-01   Residual 2.69e-02
  48   Volume ratio  8.82   Variation (σ/μ) 3.57e-01   Residual 2.23e-02
  49   Volume ratio  9.20   Variation (σ/μ) 3.58e-01   Residual 1.44e-02
  50   Volume ratio  8.95   Variation (σ/μ) 3.53e-01   Residual 1.03e-02
  51   Volume ratio  9.08   Variation (σ/μ) 3.52e-01   Residual 4.91e-03
  52   Volume ratio  9.05   Variation (σ/μ) 3.50e-01   Residual 3.98e-03
  53   Volume ratio  9.13   Variation (σ/μ) 3.51e-01   Residual 2.81e-03
  54   Volume ratio  9.11   Variation (σ/μ) 3.50e-01   Residual 2.28e-03
  55   Volume ratio  9.17   Variation (σ/μ) 3.50e-01   Residual 1.51e-03
  56   Volume ratio  9.15   Variation (σ/μ) 3.50e-01   Residual 1.11e-03
  57   Volume ratio  9.19   Variation (σ/μ) 3.50e-01   Residual 5.28e-04
  58   Volume ratio  9.18   Variation (σ/μ) 3.49e-01   Residual 2.89e-04
  59   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 1.04e-04
  60   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 7.69e-05
  61   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 5.31e-05
  62   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 3.80e-05
  63   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 2.66e-05
  64   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 1.87e-05
  65   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 1.36e-05
  66   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 9.55e-06
  67   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 7.22e-06
  68   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 5.04e-06
  69   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 3.88e-06
  70   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 2.50e-06
  71   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 1.87e-06
  72   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 9.04e-07
  73   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 6.32e-07
  74   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 2.67e-07
  75   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 1.55e-07
  76   Volume ratio  9.19   Variation (σ/μ) 3.49e-01   Residual 8.22e-08
Solver converged in 76 iterations.

Quite a few more iterations (up from 52 with np.sqrt(0.15) since it gets stuck a bit as you can see.

This also made me wonder - where did the solver parameters in movement.solver_parameters come from?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Quite a few more iterations (up from 52 with np.sqrt(0.15) since it gets stuck a bit as you can see.

Updated in a6636e0 using output from my machine. I imagine it's due to how the initial mesh is lined up, but would need to investigate in more detail. Opened a new issue: #148.

This also made me wonder - where did the solver parameters in movement.solver_parameters come from?

The ones for MA are based on the ones in Andrew McRae's 2018 paper.

)
ring_monitor = mb.get_monitor()

# With an initial mesh and a monitor function, we are able to construct a
# :class:`~movement.monge_ampere.MongeAmpereMover` instance and adapt the mesh. By
Expand Down
1 change: 1 addition & 0 deletions movement/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
from movement.monitor import * # noqa
from movement.tangling import * # noqa

from movement.laplacian_smoothing import * # noqa
Expand Down
Loading
Loading