Add BoundaryHandler and VelocityHandler #238

whzup · 2018-08-31T20:13:22Z

Description

This PR implements the new BoundaryHandler and VelocityHandler classes that deal with the issue of particles that don't move in optimisations with boundary conditions. This is achieved by using different strategies that have been described in Hel10 (P. 28). These strategies include:

Nearest bound: Reset particle to the nearest boundary
Random: Reset the particle to a random position within the bounds
Shrink: Shrink the velocity such that the particle touches the boundary instead of surpassing it
Intermediate: Reset the particle to a value between its former location and the bound
Resample: Calculate new velocities until the next particle position is within the bounds
Periodic: Tile the space with the same search space over and over

There is one more:

Reflective: Mirror the position of the particle at the surpassed boundary

I won't include the last one in this PR as it is probably provided by @jolayfield. After they've made their PR I'm going to embed their code in the BoundaryHandler class.
The strategies for the VelocityHandler include:

Unmodified: The velocity remains without modification
Adjust: The velocity is adjusted to be exactly the distance vector between the previous and the current position.
Invert: The velocity is inverted.
Zero: The velocities of particles that go out-of-bounds are set to zero.

Related Issue

Resolves: #237
See also: #150, #234

Motivation and Context

In #234 we encountered a weird phenomenon when optimizing with boundary conditions. A lot of particles stopped moving while the optimization was still running. This is caused by the waiving of updating the position in the compute_position() function when it leaves the boundaries. To fix this issue we introduce a new class called BoundaryHandler as well as a VelocityHandler which are going to reposition the particle instead of not updating it.

How Has This Been Tested?

Types of changes

Bug fix (a non-breaking change which fixes an issue)
New feature (a non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to change)

Checklist:

My code follows the code style of this project.
My change requires a change to the documentation.
I have updated the documentation accordingly.
I have read the CONTRIBUTING document.
I have added tests to cover my changes.
All new and existing tests pass.

TODO:

Reference: #198 Inheriting from a metaclass, rather than an object, grants us more flexibility and stronger typechecking for our abstract/base classes. It's also the more Pythonic option. I also added `@abs.abstractmethod` decorators for abstract methods. Signed-off-by: ljvmiranda921 <[email protected]>

Removed the `environments` module in favor of the `plotters` module. Signed-off by: ljvmiranda921 <[email protected]>

Resolves: #201 Added fmt: off and fmt: on commments to n-dimensional arrays in the test suite to prevent the ugly wrapping of arrays. Ran the black package to format the whole project.

Resolves #209 This commit adds a reporter module that implements a Reporter class. Its goal is to separate the concern of logging and printing from the optimizer classes. Signed-off-by: Lester James V. Miranda <[email protected]>

Resolves #228 Added a cost function decorator which allows the user to write a cost function that serves as a model for how the cost will be computed for every one particle. The decorator is tested with a pytest file where the shape and the equality with an example function is checked. I added a note to the documentation that some numpy functions will return arrays with single values in them.

This commit exposes the "limits" parameter in the README, and updates some undocumented parameters in the docstrings. Signed-off-by: Lester James V. Miranda <[email protected]>

Created the BoundHandler class and implemented the "nearest" strategy.

Added a whole lot of documentation and implemented the Random strategy for the position handling.

whzup · 2018-09-02T10:43:53Z

I think we should put the Resample and the Shrink strategy into the VelocityHandler as they require, unlike the others, the velocity to fix the particles position.

Added RST files.

ljvmiranda921

I think the API looks nice and is similar to how I imagined it. My only comment is that we lessen "side-effects" in the handlers: just let them return the new position, do not update self.position.

Perhaps we can pass position, bounds, *args, and **kwargs to each of the handlers; this removes the need for self.position and any other attributes. Might also make our API more generalizable given weird handlers.

I think we should put the Resample and the Shrink strategy into the VelocityHandler as they require, unlike the others, the velocity to fix the particles position.

But they still handle the position wrt bounds, right? They just need the velocity matrix, but they don't update it right? This would still be better in BoundaryHandler, let's just require velocity as one of the *args.

pyswarms/backend/handlers.py

whzup · 2018-09-03T18:47:22Z

I'm just wondering if we need the VelocityHandler if we already handle the position. Shrink and Resample use the velocity so if we handle both the velocity and the position and the same time we do things twice if not thrice. I don't see a difference between repairing particle positions after they moved outside the bounds and making the same reparations just before the particle exceeds the boundaries. The velocity handlings in the dissertation seem very similar to the position handlings we will implement.

ljvmiranda921 · 2018-09-04T00:31:14Z

I think the reason why the author adds a set of velocity handlers is that we also want to update the velocity matrix that caused these particles to be out-of-bounds. If they are all Unmodified, there's a chance that they'll go out-of-bounds again. Sure, we can call BoundaryHandler again, but it's better to pull the particles back inside the specified limits

ljvmiranda921 · 2018-09-08T10:24:50Z

Hey @whzup , I can help you with this if you want

whzup · 2018-11-29T08:17:03Z

Note: I'm going to make more or less trivial tests, only to see if the strategies do what they should do. This in combination with the test for exceeding particles should be enough for now.

Now all the strategies return the particles into the bounds

ljvmiranda921 · 2019-01-28T15:24:57Z

Hey @whzup , this honestly seems to be a difficult problem hmmm...

Fixed a bug in the topologies where all the topologie would calculate the global best position instead of the best positions of the neighbourhoods. Fixed the VonNeumann tests as well as the optimizers such that they return the position with the best cost. I had to change the parameters for the optimization in the test_objective_func_with_kwargs to satisfy the conditions. Adjusted the code to fit the refactoring of the code. Especially in the topology tests where the sytax was wrong. Fixed the test by just checking if either the best or the second best position is inside of the matrix of the best positions... Fixed the error with the class passing by adding parenthesis which seem to have disappeared Cleaned up the PR and skipped the tests which require a very good cost as some topologies do not converge fast enough. Resolves #250

For modern python, let's support py35 above Signed-off-by: Lester James V. Miranda <[email protected]>

whzup · 2019-01-31T09:22:39Z

@ljvmiranda921, What do you think, shall I try to make some tests by calculating the desired positions by hand according to the algorithms or are the current tests sufficient?

ljvmiranda921 · 2019-02-06T14:07:12Z

Hi @whzup , I think the current tests for now are sufficient. If this is good then just ping me and we can merge this to master.

Fixed some spelling mistake and added the description for the reflective strategy.

ljvmiranda921 · 2019-02-09T03:48:07Z

Merged! Thank you for your great contribution, @whzup !

ljvmiranda921 and others added 11 commits August 14, 2018 19:58

Remove environments module (#217)

ad86b71

Removed the `environments` module in favor of the `plotters` module. Signed-off by: ljvmiranda921 <[email protected]>

Remove _func in single_obj function names (#222)

2472ed2

Merge branch 'master' into development

eb05142

Merge branch 'master' into development

85952da

Add fmt comments for black (#229)

f9dc37c

Resolves: #201 Added fmt: off and fmt: on commments to n-dimensional arrays in the test suite to prevent the ugly wrapping of arrays. Ran the black package to format the whole project.

Add reporter module (#227)

0ad254d

Resolves #209 This commit adds a reporter module that implements a Reporter class. Its goal is to separate the concern of logging and printing from the optimizer classes. Signed-off-by: Lester James V. Miranda <[email protected]>

Fix plotters documentation (#236)

5961ea4

This commit exposes the "limits" parameter in the README, and updates some undocumented parameters in the docstrings. Signed-off-by: Lester James V. Miranda <[email protected]>

[WIP] Create BoundHandler class

0166cb4

Created the BoundHandler class and implemented the "nearest" strategy.

[WIP] Add documentation and implement Random

91422ed

Added a whole lot of documentation and implemented the Random strategy for the position handling.

whzup added enhancement Feature requests documentation Documentation improvements or fixes unit tests Test-related changes v.1.1.0 In pipeline for next version labels Aug 31, 2018

whzup self-assigned this Aug 31, 2018

[WIP] Add RST files for the handlers module

1b3eca0

Added RST files.

ljvmiranda921 reviewed Sep 3, 2018

View reviewed changes

ljvmiranda921 added 4 commits September 8, 2018 19:35

[WIP] Change BoundaryHandler inheritance

d9e0427

[WIP] Use dict instead of if-else for strats

6e37366

[ci skip] [WIP] Update random and nearest

1f24922

[WIP] Fix import problem in Reporter

89d1818

ljvmiranda921 self-assigned this Sep 8, 2018

[WIP] Automatically generate strategies dict

ab6aadb

[WIP] Fix all the Boundary handling strategies

e04cf15

Now all the strategies return the particles into the bounds

whzup mentioned this pull request Dec 14, 2018

(De-)Coupling between operators and topologies #259

Closed

whzup added 4 commits December 15, 2018 18:34

[WIP] Introduce the handlers into the optimizers

b7885d4

[WIP] Improve documentation

fd7ec98

[WIP] Reformat changed files

8669e6d

[WIP] Implement reflective strategy

40becad

whzup and others added 9 commits January 28, 2019 23:32

Add new targets for Makefile

2f56cf5

Use venv instead of env

ff4558e

Use pip-tools for creating dependencies

48b8042

Update setup.py to read directly from requirements

20e88a9

Add requirements files to MANIFEST.in

c5a5157

Remove py34 support

77e3fcd

For modern python, let's support py35 above Signed-off-by: Lester James V. Miranda <[email protected]>

Update HISTORY.rst

2bafd39

Merge branch 'development' into bound-handling

5fa4fa7

whzup changed the title ~~[WIP] Add BoundaryHandler class~~ [WIP] Add BoundaryHandler and VelocityHandler Jan 31, 2019

whzup changed the title ~~[WIP] Add BoundaryHandler and VelocityHandler~~ Add BoundaryHandler and VelocityHandler Feb 7, 2019

Fix and update documentation for handlers

a1127ed

Fixed some spelling mistake and added the description for the reflective strategy.

ljvmiranda921 changed the base branch from development to master February 9, 2019 03:19

Merge branch 'master' into bound-handling

fa4a2eb

ljvmiranda921 merged commit 50f6e15 into ljvmiranda921:master Feb 9, 2019

This was referenced Feb 14, 2019

Handlers features from #238 removed duplicate calls fixes from #266 #289

Closed

Re-fix duplicate cost function calls in optimize() #290

Merged

whzup deleted the bound-handling branch February 22, 2019 19:46

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add BoundaryHandler and VelocityHandler #238

Add BoundaryHandler and VelocityHandler #238

whzup commented Aug 31, 2018 •

edited

Loading

whzup commented Sep 2, 2018

ljvmiranda921 left a comment •

edited

Loading

whzup commented Sep 3, 2018

ljvmiranda921 commented Sep 4, 2018

ljvmiranda921 commented Sep 8, 2018

whzup commented Nov 29, 2018

ljvmiranda921 commented Jan 28, 2019

whzup commented Jan 31, 2019

ljvmiranda921 commented Feb 6, 2019

ljvmiranda921 commented Feb 9, 2019

Add BoundaryHandler and VelocityHandler #238

Add BoundaryHandler and VelocityHandler #238

Conversation

whzup commented Aug 31, 2018 • edited Loading

Description

Related Issue

Motivation and Context

How Has This Been Tested?

Types of changes

Checklist:

TODO:

whzup commented Sep 2, 2018

ljvmiranda921 left a comment • edited Loading

Choose a reason for hiding this comment

whzup commented Sep 3, 2018

ljvmiranda921 commented Sep 4, 2018

ljvmiranda921 commented Sep 8, 2018

whzup commented Nov 29, 2018

ljvmiranda921 commented Jan 28, 2019

whzup commented Jan 31, 2019

ljvmiranda921 commented Feb 6, 2019

ljvmiranda921 commented Feb 9, 2019

whzup commented Aug 31, 2018 •

edited

Loading

ljvmiranda921 left a comment •

edited

Loading