workbench for performing exploratory modeling and analysis
BSD-3-CLAUSE License
Bot releases are hidden (Show)
2.5.2 is a small patch release that makes the EMAworkbench compatible with NumPy 2.0 and improves the MPIEvaluator tutorial.
Full Changelog: https://github.com/quaquel/EMAworkbench/compare/2.5.1...2.5.2
Published by EwoutH 6 months ago
The 2.5.1 release is a small patch release with two bugfixes.
futures_util.py
incorrectly assumed the presence of experiment_runner
in its module namespace, leading to failures in futures models like multiprocessing and mpi. This is resolved by adjusting the finalizer function to expect experiment_runner
as an argument.Full Changelog: https://github.com/quaquel/EMAworkbench/compare/2.5.0...2.5.1
Published by EwoutH 10 months ago
In the 2.5.0 release of the EMAworkbench we introduce a new experimental MPIevaluator to run on multi-node (HPC) systems (#299, #328). We would love feedback on it in #311.
Furthermore, the pair plots for scenario discovery now allow contour plots and bivariate histograms (#288). When doing Prim you can inspect multiple boxes and display them in a single figure (#317).
From 3.0 onwards, the names of parameters, constants, constraints, and outcomes must be valid python identifiers. From this version onwards, a DeprecationWarning is raised if the name is not a valid Python identifier.
Full Changelog: https://github.com/quaquel/EMAworkbench/compare/2.4.0...2.5.0
Published by EwoutH over 1 year ago
2.4.1 is a small patch release of the EMAworkbench, primarily resolving issues #276 and #277 in the workbench itself, and a bug introduced by #241 in the docs. The EMAworkbench now also raise exception when sampling scenarios or policies while no uncertainties or levers are defined (#282).
rebuild_platypus_population
by @quaquel in https://github.com/quaquel/EMAworkbench/pull/276
load_results
function. The dtype metadata is now correctly applied, resolving issue #277.n_processes=-1
in MultiprocessingEvaluator by @EwoutH in https://github.com/quaquel/EMAworkbench/pull/278
Full Changelog: https://github.com/quaquel/EMAworkbench/compare/2.4.0...2.4.1
Published by quaquel over 1 year ago
The latest release of the EMAworkbench introduces significant performance improvements and quality of life updates. The performance of _store_outcomes
has been enhanced by approximately 35x in pull request #232, while the combine
function has seen a 8x speedup in pull request #233. This results in the overhead of the EMAworkbench being reduced by over 70%. In a benchmark, a very simple Python model now performs approximately 40.000 iterations per second, compared to 15.000 in 2.3.0.
In addition to these performance upgrades, the examples have been added to the ReadTheDocs documentation, more documentation improvements have been made and many bugs and deprecations have been fixed.
The 2.4.x release series requires Python 3.8 and is tested on 3.8 to 3.11. It can be installed as usual via PyPI, with:
pip install --upgrade ema-workbench
combine
by ~8x by @EwoutH in https://github.com/quaquel/EMAworkbench/pull/233
pynetlogo
imports by decapitalizing pyNetLogo
by @quaquel in https://github.com/quaquel/EMAworkbench/pull/248
metric
argument to allow for other linkages by @mikhailsirenko in https://github.com/quaquel/EMAworkbench/pull/222
Full Changelog: https://github.com/quaquel/EMAworkbench/compare/2.3.0...2.4.0
Published by quaquel almost 2 years ago
This release adds a new algorithm for output space exploration. The way in which convergence tracking for optimization is supported has been overhauled completely, see the updated directed search user guide for full details. The documentation has moreover been expanded with a comparison to Rhodium.
With this new release, the installation process has been improved by reducing the number of required dependencies. Recommended packages and connectors can now be installed as extras using pip, for example pip install -U ema-workbench[recommended,netlogo]
. See the updated installation instructions for all options and details.
The 2.3.x release series supports Python 3.8 to 3.11. It can be installed as usual via PyPI, with:
pip install --upgrade ema-workbench
n_processes
input in MultiprocessingEvaluator by @EwoutH in https://github.com/quaquel/EMAworkbench/pull/189
_optimize()
by @EwoutH in https://github.com/quaquel/EMAworkbench/pull/150
shade
keyword in Seaborn kdeplot with fill
by @EwoutH in https://github.com/quaquel/EMAworkbench/pull/169
saltelli
with sobol
SALib 1.4.6+. by @quaquel in https://github.com/quaquel/EMAworkbench/pull/211
Full Changelog: https://github.com/quaquel/EMAworkbench/compare/2.2.0...2.3.0
Published by EwoutH about 2 years ago
The 2.2.1 release is a small patch release to address 3 bugs in the EMAworkbench 2.2.0 (#155, #168 and #188). Thanks @vioSpark and @steipatr for reporting those!
This new version can be installed as usual via PyPI, with:
pip install --upgrade ema-workbench
n_processes
input in MultiprocessingEvaluator by @EwoutH in https://github.com/quaquel/EMAworkbench/pull/189
Full Changelog: https://github.com/quaquel/EMAworkbench/compare/2.2.0...2.2.1
Published by EwoutH about 2 years ago
With the 2.2 release, the EMAworkbench can now connect to Vadere models on pedestrian dynamics. When inspecting a Prim Box peeling trajectory, multiple points on the peeling trajectory can be inspected simultaneously by inputting a list of integers into PrimBox.inspect()
.
When running experiments with multiprocessing using the MultiprocessingEvaluator
, the number of processes can now be controlled using a negative integer as input for n_processes
(for example, -2
on a 12-thread CPU results in 10 threads used). Also, it will now default to max. 61 processes on windows machines due to limitations inherent in Windows in dealing with higher processer counts. Code quality, CI, and error reporting also have been improved. And finally, generating these release notes is now automated.
This new version can be installed as usual via PyPI, with:
pip install --upgrade ema-workbench
PrimBox.peeling_trajectory["id"]
int instead of float by @EwoutH in https://github.com/quaquel/EMAworkbench/pull/121
PrimBox.inspect()
now can also take a list of integers (aside from a single int) to inspect multiple points at once by @quaquel in https://github.com/quaquel/EMAworkbench/commit/6d83a6c33442ad4dce0974a384b03a225aaf830d (see also issue https://github.com/quaquel/EMAworkbench/issues/124)max_features
in get_rf_feature_scores to "sqrt"
by @EwoutH in https://github.com/quaquel/EMAworkbench/pull/129
project_urls
for documentation and issue tracker links by @EwoutH in https://github.com/quaquel/EMAworkbench/pull/142
Full Changelog: https://github.com/quaquel/EMAworkbench/compare/2.1.2...2.2.0
Published by quaquel over 5 years ago
2.0 version of the workbench.
Main changes:
Published by quaquel about 9 years ago
Stable version of the workbench