[](https://badge.fury.io/py/coexist)
[](https://coexist.readthedocs.io/en/latest/?badge=latest)
[](https://github.com/uob-positron-imaging-centre/Coexist/actions/workflows/ci.yml)
[](https://lgtm.com/projects/g/uob-positron-imaging-centre/Coexist/context:python)
[](https://colab.research.google.com/drive/1p7OwnaYgENwK4DTn_6QETX4ajwVFeza0?usp=sharing)
[](https://github.com/uob-positron-imaging-centre/Coexist)
# CoExSiST & ACCES
### Data-Driven Evolutionary Calibration & Optimisation of Simulations
A Python library for autonomously learning simulation parameters from experimental data, from the *micro* to the *macro*, from laptops to clusters. This is done using either of two closely related frameworks:
- **CoExSiST**: Coupled Experimental-Simulational Study Tool.
- **ACCES**: Autonomous Characterisation and Calibration via Evolutionary Simulation.
Both libraries learn a given set of free parameters, such that an experiment is synchronised with an equivalent simulation; this synchronisation is done in one of two ways:
- CoExSiST calibrates **microscopically**: in a Discrete Element Method (DEM) context, all simulated particles follow their experimental counterparts *exactly*. Naturally, this technique is limited to dilute systems and experimental imaging techniques that can capture the 3D position of *all* moving particles (e.g. PIV) - however, it provides information about the fundamental aspects of particle collision.
- ACCES calibrates / optimises **macroscopically**: a given simulation reproduces a system-specific *macroscopic* quantity (e.g. residence time distribution, angle of repose). This technique is completely agnostic to the simulation method and the quantity to be reproduced. For example, it can train *coarse-grained DEM* simulations, using larger meso-particles to model multiple smaller ones.
ACCES is ready for production use; it was successfully used to calibrate coarse-grained DEM digital twins of [GranuTools](https://www.granutools.com/en/) equipment (Andrei Leonard Nicusan and Dominik Werner, *paper under review*), CFDEM fluidised beds (Hanqiao Cha, *paper under review*) and even signal processing parameters in a PET scanner model (Matthew Herald, *paper under review*).
*Example of an ACCES-calibrated DEM Digital Twin of a GranuTools GranuDrum; the calibration was done effectively against a single experimental data point - a photograph of the free surface shape yielded by MCC particles (left panel). The occupancy grid of a LIGGGHTS simulation was optimised against the free surface shape (right panel). The two superimposed grids amount to 4 mm² dissimilarity (dark blue pixels, middle panel).*
ACCES was implemented in the `coexist.Access` class, providing an interface that is easy to use, but powerful enough to **automatically parallelise arbitrary Python scripts** through code inspection and metaprogramming. It was used successfully from laptop-scale shared-memory machines to multi-node supercomputing clusters.
## Getting Started
This is a pure Python package that does not require any extra system configuration, supporting Python 3.6 and above (though it might work with even older versions) - to install it from PyPI, simply run:
```
pip install coexist
```
Or you can install the development version from the GitHub repository:
```
pip install git+https://github.com/uob-positron-imaging-centre/Coexist
```
### Examples
The [documentation](https://coexist.readthedocs.io/) website contains an ACCES [tutorial](https://coexist.readthedocs.io/en/latest/tutorials/index.html) with explained code and output figures produced by `coexist`; all public functionality is fully documented in the [manual](https://coexist.readthedocs.io/en/latest/manual/index.html).
Want something more hands on? Check out the `examples` folder for example scripts using `coexist.Coexist` and `coexist.Access`; `examples/access_simple` is a very simple, hackable example script (remember that the `simulation_script.py` can execute *anything*). For a more involved, complete calibration of a GranuTools GranuDrum digital twin - using LIGGGHTS - see our collection of peer-reviewed digital twins [repository](https://github.com/uob-positron-imaging-centre/DigitalTwins).
The `coexist.plots` submodule can plot the convergence of ACCES onto optimally-calibrated parameters - or check intermediate results while your 50-hour simulations runs:
### Show me some code already...
Alright, here's the gist of it: instead of having to rewrite your complex simulation into a function for calibration / optimisation (which is what virtually all optimisation frameworks require...), ACCES takes in an _entire simulation script_; here's a simple example, say in a file called `simulation_script.py`:
```python
# File: simulation_script.py
# ACCESS PARAMETERS START
import coexist
parameters = coexist.create_parameters(
variables = ["CED", "CoR", "Epsilon", "Mu"],
minimums = [-5, -5, -5, -5],
maximums = [+5, +5, +5, +5],
values = [0, 0, 0, 0], # Optional, initial guess
)
access_id = 0 # Optional, unique ID for each simulation
# ACCESS PARAMETERS END
# Extract the free parameters' values - ACCES will modify / optimise them.
x, y, z, t = parameters["value"]
# Define the error value in *any* way - run a simulation, analyse data, etc.
#
# For simplicity, here's an analytical 4D Himmelblau function with 8 global
# and 2 local minima - the initial guess is very close to the local one!
error = (x**2 + y - 11)**2 + (x + y**2 - 7)**2 + z * t
```
Then you can run in a separate script or Python console:
```python
# File: access_learn.py
import coexist
# Use ACCESS to learn a simulation's parameters
access = coexist.Access("simulation_script.py")
access.learn(
num_solutions = 10, # Number of solutions per epoch
target_sigma = 0.1, # Target std-dev (accuracy) of solution
random_seed = 42, # Reproducible / deterministic optimisation
)
# Or simply
# coexist.Access("simulation_script.py").learn(random_seed = 42)
```
That's it - ACCES will take the `simulation_script.py`, modify the free `parameters`, run the script in parallel (either on all processors of your local machine or on a distributed supercomputer) and optimise the `error` you defined; it'll look something like this:
```
================================================================================
Starting ACCES run at 08:32:23 on 01/02/2022 in directory `access_seed42`.
================================================================================
Epoch 0 | Population 10
--------------------------------------------------------------------------------
Scaled overall standard deviation: 1.0
CED CoR Epsilon Mu
estimate 0.0 0.000000 0.000000 0.000000
uncertainty 4.0 4.000050 4.000100 4.000150
scaled_std 1.0 1.000013 1.000025 1.000038
CED CoR Epsilon Mu error
0 1.218868 -4.159988 3.001880 3.762400 331.093228
1 -3.095859 -4.967675 0.511374 -1.265018 252.732762
2 -0.067205 -3.412218 3.517680 3.111284 239.466423
3 0.264123 4.509021 1.870084 -3.437299 219.638764
4 1.475003 -3.835578 3.513889 -0.199711 243.967225
5 -0.739449 -2.723752 4.825957 -0.618141 170.752129
6 -1.713311 -1.408552 2.129290 1.461831 138.135979
7 1.650930 1.723306 2.333195 -1.625721 44.785098
8 -2.048971 -3.255132 2.463979 4.516059 114.660635
9 -0.455790 -3.360668 -3.298007 2.602469 206.454823
Total function evaluations: 10
================================================================================
Epoch 1 | Population 10
--------------------------------------------------------------------------------
...
<output truncated>
...
================================================================================
Epoch 31 | Population 10
--------------------------------------------------------------------------------
Scaled overall standard deviation: 0.1032642976669169
CED CoR Epsilon Mu
estimate 3.621098 -1.748473 4.999445 -4.992881
uncertainty 0.052355 0.078080 0.284105 0.180196
scaled_std 0.013089 0.019520 0.071026 0.045049
CED CoR Epsilon Mu error
310 3.574473 -1.650134 4.999374 -4.901913 -23.996813
311 3.582026 -1.756384 4.863496 -4.973417 -24.071692
312 3.628789 -1.616891 4.859678 -4.992656 -23.386000
313 3.662351 -1.782704 4.982040 -4.998059 -24.478008
314 3.594971 -1.725715 4.999877 -4.898257 -24.269162
315 3.577725 -1.744971 4.998411 -4.956502 -24.629198
316 3.613914 -1.747412 4.983253 -4.996630 -24.690880
317 3.579212 -1.774811 4.852262 -4.972910 -24.055219
318 3.634952 -1.784927 4.999971 -4.999863 -24.783959
319 3.647169 -1.872419 4.999971 -4.978640 -24.685207
Total function evaluations: 320
Optimal solution found within `target_sigma`, i.e. 10.0%:
sigma = 0.08390460663313491 < 0.1
================================================================================
The best result was found in 32 epochs:
CED CoR Epsilon Mu error
value 3.569249 -1.813354 4.995112 -4.994052 -24.920092
std 0.042168 0.060116 0.203900 0.140874
These results were found for the job:
access_seed42/results/parameters.284.pickle
================================================================================
```
And you can access (pun intended) the results - even as ACCES is running - using:
```python
>>> import coexist
>>> data = coexist.AccessData.read("access_seed42")
>>> data
AccessData
--------------------------------------------------------------------------------
paths ╎ AccessPaths(...)
parameters ╎ value min max sigma
╎ CED 3.569249 -5.0 5.0 0.052355
╎ CoR -1.813354 -5.0 5.0 0.078080
╎ Epsilon 4.995112 -5.0 5.0 0.284105
╎ Mu -4.994052 -5.0 5.0 0.180196
population ╎ 10
num_epochs ╎ 32
target ╎ 0.1
seed ╎ 42
epochs ╎ DataFrame(CED_mean, CoR_mean, Epsilon_mean, Mu_mean, CED_std,
╎ CoR_std, Epsilon_std, Mu_std, overall_std)
epochs_scaled ╎ DataFrame(CED_mean, CoR_mean, Epsilon_mean, Mu_mean, CED_std,
╎ CoR_std, Epsilon_std, Mu_std, overall_std)
results ╎ DataFrame(CED, CoR, Epsilon, Mu, error)
results_scaled ╎ DataFrame(CED, CoR, Epsilon, Mu, error)
```
In this case a global optimum was found within 320 evaluations - this is of course problem-dependent, but you'll see that the optimum is often found much earlier if you check intermediate results (which you probably will when calibrating / optimising long-running simulations).
A tutorial with more detailed explanations is available [here](https://coexist.readthedocs.io/en/latest/tutorials/index.html), including generating plots, checking intermediate results and running simulations on a SLURM-managed distributed cluster.
## Contributing
This library aims to be the state-of-the-art for simulation calibration, developed in the open using modern, collaborative coding approaches - no dragons shall be dwelling in the codebase. You are more than welcome to contribute to this library in the form of code improvements, documentation or helpful examples; please submit them either as:
- GitHub issues.
- Pull requests.
- Email me at <a.l.nicusan@gmail.com>.
We are more than happy to discuss the library architecture and calibration / optimisation approach with any potential contributors and users.
## Acknowledgements and Funding
The authors gratefully acknowledge funding from the following UK funding bodies and industrial partners:
**M²E³D: Multiphase Materials Exploration via Evolutionary Equation Discovery**
Royce Materials 4.0 Feasibility and Pilot Scheme Grant, £57,477
**CoExSiST: Coupled Experimental-Simulational Technique**
EPSRC MAPP Grant, Feasibility Study, £60,246
**ACCES: Autonomous Calibration and Characterisation via Evolutionary Simulation**
EPSRC IAA, Follow-Up Grant to CoExSiST, £52,762
**Improving ACCES: Towards the Multi-Tool Multi-Parameter Optimisation of Complex Particulate Systems**
EPSRC MAPP, Grant, £48,871 + £48,871 matched funding from GranuTools Belgium
Thank you.
## Citing
If you use this library in your research, you are kindly asked to cite:
> [Paper after publication]
Until the ACCES paper is published, you may cite this repository:
> Nicusan AL, Werner D, Sykes J, Seville JPK, Windows-Yule CR. ACCES: Autonomous Characterisation and Calibration via Evolutionary Simulation. GitHub repository. 2022 February 1.
ACCES is built on top of the excellent CMA-ES evolutionary algorithm - specifically the [`pycma`](https://github.com/CMA-ES/pycma) implementation. If you use ACCES in your research, please also cite:
> Nikolaus Hansen, Youhei Akimoto, and Petr Baudis. CMA-ES/pycma on Github. Zenodo, DOI:10.5281/zenodo.2559634, February 2019.
## License and Commercial Integration
This library - in its general, domain-agnostic form - is free and open-source, published under the GNU General Public License v3.0.
If you are a company and would like to integrate ACCESS into your work - e.g. ACCESS-enabled equipment or general simulation calibration - please send an email to `a.l.nicusan@bham.ac.uk` to discuss commercial development of specific tools for your application. Relicensing for a closed-source / commercial project can be considered on an individual basis.
Copyright (C) 2020-2023 the Coexist developers. Until now, this library was built directly or indirectly through the brain-time of:
- Andrei Leonard Nicusan (University of Birmingham)
- Dominik Werner (University of Birmingham)
- Jack Sykes (University of Birmingham)
- Dr. Kit Windows-Yule (University of Birmingham)
- Prof. Jonathan Seville (University of Birmingham)
- Albert Bauer (TU Berlin)
Thank you.
Raw data
{
"_id": null,
"home_page": "https://github.com/uob-positron-imaging-centre/Coexist",
"name": "coexist",
"maintainer": "",
"docs_url": null,
"requires_python": ">=3.6",
"maintainer_email": "",
"keywords": "simulation optimization calibration parameter-estimation parameter-tuning physics-simulation computational-fluid-dynamics granular DEM discrete-element-method medical-imaging",
"author": "Andrei Leonard Nicusan <a.l.nicusan@bham.ac.uk>, Dominik Werner <d.werner.1@bham.ac.uk>, Jack Sykes <jas653@student.bham.ac.uk>",
"author_email": "",
"download_url": "https://files.pythonhosted.org/packages/b7/c8/f2b192f0729cef9f96686069638bfb857dcb431eceeeecc9d668f6b65f90/coexist-0.3.2.tar.gz",
"platform": null,
"description": "[](https://badge.fury.io/py/coexist)\n[](https://coexist.readthedocs.io/en/latest/?badge=latest)\n[](https://github.com/uob-positron-imaging-centre/Coexist/actions/workflows/ci.yml)\n[](https://lgtm.com/projects/g/uob-positron-imaging-centre/Coexist/context:python)\n[](https://colab.research.google.com/drive/1p7OwnaYgENwK4DTn_6QETX4ajwVFeza0?usp=sharing)\n[](https://github.com/uob-positron-imaging-centre/Coexist)\n\n\n# CoExSiST & ACCES\n### Data-Driven Evolutionary Calibration & Optimisation of Simulations\n\nA Python library for autonomously learning simulation parameters from experimental data, from the *micro* to the *macro*, from laptops to clusters. This is done using either of two closely related frameworks:\n\n- **CoExSiST**: Coupled Experimental-Simulational Study Tool.\n- **ACCES**: Autonomous Characterisation and Calibration via Evolutionary Simulation. \n\nBoth libraries learn a given set of free parameters, such that an experiment is synchronised with an equivalent simulation; this synchronisation is done in one of two ways:\n\n- CoExSiST calibrates **microscopically**: in a Discrete Element Method (DEM) context, all simulated particles follow their experimental counterparts *exactly*. Naturally, this technique is limited to dilute systems and experimental imaging techniques that can capture the 3D position of *all* moving particles (e.g. PIV) - however, it provides information about the fundamental aspects of particle collision.\n- ACCES calibrates / optimises **macroscopically**: a given simulation reproduces a system-specific *macroscopic* quantity (e.g. residence time distribution, angle of repose). This technique is completely agnostic to the simulation method and the quantity to be reproduced. For example, it can train *coarse-grained DEM* simulations, using larger meso-particles to model multiple smaller ones.\n\n\nACCES is ready for production use; it was successfully used to calibrate coarse-grained DEM digital twins of [GranuTools](https://www.granutools.com/en/) equipment (Andrei Leonard Nicusan and Dominik Werner, *paper under review*), CFDEM fluidised beds (Hanqiao Cha, *paper under review*) and even signal processing parameters in a PET scanner model (Matthew Herald, *paper under review*).\n\n\n![Calibrated GranuDrum](/docs/source/_static/calibrated.png?raw=true \"Calibrated GranuDrum.\")\n*Example of an ACCES-calibrated DEM Digital Twin of a GranuTools GranuDrum; the calibration was done effectively against a single experimental data point - a photograph of the free surface shape yielded by MCC particles (left panel). The occupancy grid of a LIGGGHTS simulation was optimised against the free surface shape (right panel). The two superimposed grids amount to 4 mm\u00b2 dissimilarity (dark blue pixels, middle panel).*\n\n\nACCES was implemented in the `coexist.Access` class, providing an interface that is easy to use, but powerful enough to **automatically parallelise arbitrary Python scripts** through code inspection and metaprogramming. It was used successfully from laptop-scale shared-memory machines to multi-node supercomputing clusters.\n\n\n\n\n\n## Getting Started\n\nThis is a pure Python package that does not require any extra system configuration, supporting Python 3.6 and above (though it might work with even older versions) - to install it from PyPI, simply run:\n\n```\npip install coexist\n```\n\nOr you can install the development version from the GitHub repository:\n\n```\npip install git+https://github.com/uob-positron-imaging-centre/Coexist\n```\n\n\n\n\n### Examples\n\nThe [documentation](https://coexist.readthedocs.io/) website contains an ACCES [tutorial](https://coexist.readthedocs.io/en/latest/tutorials/index.html) with explained code and output figures produced by `coexist`; all public functionality is fully documented in the [manual](https://coexist.readthedocs.io/en/latest/manual/index.html).\n\nWant something more hands on? Check out the `examples` folder for example scripts using `coexist.Coexist` and `coexist.Access`; `examples/access_simple` is a very simple, hackable example script (remember that the `simulation_script.py` can execute *anything*). For a more involved, complete calibration of a GranuTools GranuDrum digital twin - using LIGGGHTS - see our collection of peer-reviewed digital twins [repository](https://github.com/uob-positron-imaging-centre/DigitalTwins).\n\n\n![GranuDrum ACCES Example](/docs/source/_static/access_example.png?raw=true \"GranuDrum ACCES Example.\")\n\n\nThe `coexist.plots` submodule can plot the convergence of ACCES onto optimally-calibrated parameters - or check intermediate results while your 50-hour simulations runs:\n\n\n![Convergence Plot](/docs/source/_static/access_convergence.png?raw=true \"ACCES Convergence Plot.\")\n\n\n\n\n\n### Show me some code already...\n\nAlright, here's the gist of it: instead of having to rewrite your complex simulation into a function for calibration / optimisation (which is what virtually all optimisation frameworks require...), ACCES takes in an _entire simulation script_; here's a simple example, say in a file called `simulation_script.py`:\n\n```python\n# File: simulation_script.py\n\n# ACCESS PARAMETERS START\nimport coexist\n\nparameters = coexist.create_parameters(\n variables = [\"CED\", \"CoR\", \"Epsilon\", \"Mu\"],\n minimums = [-5, -5, -5, -5],\n maximums = [+5, +5, +5, +5],\n values = [0, 0, 0, 0], # Optional, initial guess\n)\n\naccess_id = 0 # Optional, unique ID for each simulation\n# ACCESS PARAMETERS END\n\n\n# Extract the free parameters' values - ACCES will modify / optimise them.\nx, y, z, t = parameters[\"value\"]\n\n\n# Define the error value in *any* way - run a simulation, analyse data, etc.\n#\n# For simplicity, here's an analytical 4D Himmelblau function with 8 global\n# and 2 local minima - the initial guess is very close to the local one!\nerror = (x**2 + y - 11)**2 + (x + y**2 - 7)**2 + z * t\n```\n\nThen you can run in a separate script or Python console:\n\n```python\n# File: access_learn.py\n\nimport coexist\n\n# Use ACCESS to learn a simulation's parameters\naccess = coexist.Access(\"simulation_script.py\")\naccess.learn(\n num_solutions = 10, # Number of solutions per epoch\n target_sigma = 0.1, # Target std-dev (accuracy) of solution\n random_seed = 42, # Reproducible / deterministic optimisation\n)\n\n# Or simply\n# coexist.Access(\"simulation_script.py\").learn(random_seed = 42)\n```\n\nThat's it - ACCES will take the `simulation_script.py`, modify the free `parameters`, run the script in parallel (either on all processors of your local machine or on a distributed supercomputer) and optimise the `error` you defined; it'll look something like this:\n\n```\n================================================================================\nStarting ACCES run at 08:32:23 on 01/02/2022 in directory `access_seed42`.\n================================================================================\nEpoch 0 | Population 10\n--------------------------------------------------------------------------------\nScaled overall standard deviation: 1.0\n CED CoR Epsilon Mu\nestimate 0.0 0.000000 0.000000 0.000000\nuncertainty 4.0 4.000050 4.000100 4.000150\nscaled_std 1.0 1.000013 1.000025 1.000038\n\n CED CoR Epsilon Mu error\n0 1.218868 -4.159988 3.001880 3.762400 331.093228\n1 -3.095859 -4.967675 0.511374 -1.265018 252.732762\n2 -0.067205 -3.412218 3.517680 3.111284 239.466423\n3 0.264123 4.509021 1.870084 -3.437299 219.638764\n4 1.475003 -3.835578 3.513889 -0.199711 243.967225\n5 -0.739449 -2.723752 4.825957 -0.618141 170.752129\n6 -1.713311 -1.408552 2.129290 1.461831 138.135979\n7 1.650930 1.723306 2.333195 -1.625721 44.785098\n8 -2.048971 -3.255132 2.463979 4.516059 114.660635\n9 -0.455790 -3.360668 -3.298007 2.602469 206.454823\nTotal function evaluations: 10\n\n================================================================================\nEpoch 1 | Population 10\n--------------------------------------------------------------------------------\n...\n<output truncated>\n...\n================================================================================\nEpoch 31 | Population 10\n--------------------------------------------------------------------------------\nScaled overall standard deviation: 0.1032642976669169\n CED CoR Epsilon Mu\nestimate 3.621098 -1.748473 4.999445 -4.992881\nuncertainty 0.052355 0.078080 0.284105 0.180196\nscaled_std 0.013089 0.019520 0.071026 0.045049\n\n CED CoR Epsilon Mu error\n310 3.574473 -1.650134 4.999374 -4.901913 -23.996813\n311 3.582026 -1.756384 4.863496 -4.973417 -24.071692\n312 3.628789 -1.616891 4.859678 -4.992656 -23.386000\n313 3.662351 -1.782704 4.982040 -4.998059 -24.478008\n314 3.594971 -1.725715 4.999877 -4.898257 -24.269162\n315 3.577725 -1.744971 4.998411 -4.956502 -24.629198\n316 3.613914 -1.747412 4.983253 -4.996630 -24.690880\n317 3.579212 -1.774811 4.852262 -4.972910 -24.055219\n318 3.634952 -1.784927 4.999971 -4.999863 -24.783959\n319 3.647169 -1.872419 4.999971 -4.978640 -24.685207\nTotal function evaluations: 320\n\n\nOptimal solution found within `target_sigma`, i.e. 10.0%:\n sigma = 0.08390460663313491 < 0.1\n\n================================================================================\nThe best result was found in 32 epochs:\n CED CoR Epsilon Mu error\n value 3.569249 -1.813354 4.995112 -4.994052 -24.920092\n std 0.042168 0.060116 0.203900 0.140874\n\nThese results were found for the job:\n access_seed42/results/parameters.284.pickle\n================================================================================\n```\n\nAnd you can access (pun intended) the results - even as ACCES is running - using:\n\n```python\n>>> import coexist\n>>> data = coexist.AccessData.read(\"access_seed42\")\n>>> data\nAccessData\n--------------------------------------------------------------------------------\npaths \u254e AccessPaths(...)\nparameters \u254e value min max sigma\n \u254e CED 3.569249 -5.0 5.0 0.052355\n \u254e CoR -1.813354 -5.0 5.0 0.078080\n \u254e Epsilon 4.995112 -5.0 5.0 0.284105\n \u254e Mu -4.994052 -5.0 5.0 0.180196\npopulation \u254e 10\nnum_epochs \u254e 32\ntarget \u254e 0.1\nseed \u254e 42\nepochs \u254e DataFrame(CED_mean, CoR_mean, Epsilon_mean, Mu_mean, CED_std,\n \u254e CoR_std, Epsilon_std, Mu_std, overall_std)\nepochs_scaled \u254e DataFrame(CED_mean, CoR_mean, Epsilon_mean, Mu_mean, CED_std,\n \u254e CoR_std, Epsilon_std, Mu_std, overall_std)\nresults \u254e DataFrame(CED, CoR, Epsilon, Mu, error)\nresults_scaled \u254e DataFrame(CED, CoR, Epsilon, Mu, error)\n```\n\nIn this case a global optimum was found within 320 evaluations - this is of course problem-dependent, but you'll see that the optimum is often found much earlier if you check intermediate results (which you probably will when calibrating / optimising long-running simulations).\n\nA tutorial with more detailed explanations is available [here](https://coexist.readthedocs.io/en/latest/tutorials/index.html), including generating plots, checking intermediate results and running simulations on a SLURM-managed distributed cluster.\n\n\n\n\n## Contributing\n\nThis library aims to be the state-of-the-art for simulation calibration, developed in the open using modern, collaborative coding approaches - no dragons shall be dwelling in the codebase. You are more than welcome to contribute to this library in the form of code improvements, documentation or helpful examples; please submit them either as:\n\n- GitHub issues.\n- Pull requests.\n- Email me at <a.l.nicusan@gmail.com>.\n\nWe are more than happy to discuss the library architecture and calibration / optimisation approach with any potential contributors and users.\n\n\n\n\n## Acknowledgements and Funding\n\nThe authors gratefully acknowledge funding from the following UK funding bodies and industrial partners:\n\n**M\u00b2E\u00b3D: Multiphase Materials Exploration via Evolutionary Equation Discovery** \nRoyce Materials 4.0 Feasibility and Pilot Scheme Grant, \u00a357,477 \n\n**CoExSiST: Coupled Experimental-Simulational Technique** \nEPSRC MAPP Grant, Feasibility Study, \u00a360,246 \n\n**ACCES: Autonomous Calibration and Characterisation via Evolutionary Simulation** \nEPSRC IAA, Follow-Up Grant to CoExSiST, \u00a352,762 \n\n**Improving ACCES: Towards the Multi-Tool Multi-Parameter Optimisation of Complex Particulate Systems** \nEPSRC MAPP, Grant, \u00a348,871 + \u00a348,871 matched funding from GranuTools Belgium \n\nThank you.\n\n\n\n\n## Citing\n\nIf you use this library in your research, you are kindly asked to cite:\n\n> [Paper after publication]\n\n\nUntil the ACCES paper is published, you may cite this repository:\n\n> Nicusan AL, Werner D, Sykes J, Seville JPK, Windows-Yule CR. ACCES: Autonomous Characterisation and Calibration via Evolutionary Simulation. GitHub repository. 2022 February 1.\n\n\nACCES is built on top of the excellent CMA-ES evolutionary algorithm - specifically the [`pycma`](https://github.com/CMA-ES/pycma) implementation. If you use ACCES in your research, please also cite:\n\n> Nikolaus Hansen, Youhei Akimoto, and Petr Baudis. CMA-ES/pycma on Github. Zenodo, DOI:10.5281/zenodo.2559634, February 2019.\n\n\n\n\n## License and Commercial Integration\n\nThis library - in its general, domain-agnostic form - is free and open-source, published under the GNU General Public License v3.0.\n\nIf you are a company and would like to integrate ACCESS into your work - e.g. ACCESS-enabled equipment or general simulation calibration - please send an email to `a.l.nicusan@bham.ac.uk` to discuss commercial development of specific tools for your application. Relicensing for a closed-source / commercial project can be considered on an individual basis.\n\n\nCopyright (C) 2020-2023 the Coexist developers. Until now, this library was built directly or indirectly through the brain-time of:\n\n- Andrei Leonard Nicusan (University of Birmingham)\n- Dominik Werner (University of Birmingham)\n- Jack Sykes (University of Birmingham)\n- Dr. Kit Windows-Yule (University of Birmingham)\n- Prof. Jonathan Seville (University of Birmingham)\n- Albert Bauer (TU Berlin)\n\nThank you.\n",
"bugtrack_url": null,
"license": "GNU",
"summary": "Learning simulation parameters from experimental data, from the micro to the macro, from the laptop to the cluster.",
"version": "0.3.2",
"project_urls": {
"Homepage": "https://github.com/uob-positron-imaging-centre/Coexist"
},
"split_keywords": [
"simulation",
"optimization",
"calibration",
"parameter-estimation",
"parameter-tuning",
"physics-simulation",
"computational-fluid-dynamics",
"granular",
"dem",
"discrete-element-method",
"medical-imaging"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "8146bdb21d6ccf75666af516cf7020f7dfcdf91785e75b1f5ee107e221401c45",
"md5": "d779dfcfc9d0cda537c1ed4f1523c928",
"sha256": "47b09e63847220bd4a577edf9eb725cadf0c089ef535219ed7ea2bae17001c61"
},
"downloads": -1,
"filename": "coexist-0.3.2-py2.py3-none-any.whl",
"has_sig": false,
"md5_digest": "d779dfcfc9d0cda537c1ed4f1523c928",
"packagetype": "bdist_wheel",
"python_version": "py2.py3",
"requires_python": ">=3.6",
"size": 82987,
"upload_time": "2023-06-11T02:52:55",
"upload_time_iso_8601": "2023-06-11T02:52:55.420236Z",
"url": "https://files.pythonhosted.org/packages/81/46/bdb21d6ccf75666af516cf7020f7dfcdf91785e75b1f5ee107e221401c45/coexist-0.3.2-py2.py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "b7c8f2b192f0729cef9f96686069638bfb857dcb431eceeeecc9d668f6b65f90",
"md5": "d9175cc580f5c141e0c6faf1a7afd821",
"sha256": "c82d659de5d9a2ebeea7490166f44dd1d3e5788dcd831c0940a04b9eb124dd36"
},
"downloads": -1,
"filename": "coexist-0.3.2.tar.gz",
"has_sig": false,
"md5_digest": "d9175cc580f5c141e0c6faf1a7afd821",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.6",
"size": 86801,
"upload_time": "2023-06-11T02:52:57",
"upload_time_iso_8601": "2023-06-11T02:52:57.464611Z",
"url": "https://files.pythonhosted.org/packages/b7/c8/f2b192f0729cef9f96686069638bfb857dcb431eceeeecc9d668f6b65f90/coexist-0.3.2.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2023-06-11 02:52:57",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "uob-positron-imaging-centre",
"github_project": "Coexist",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"requirements": [],
"lcname": "coexist"
}
JSON
