# BOCoDe: Benchmarks for Optimization and Computational Design
We present BOCoDe, a Python and PyTorch-based library that contains the most comprehensive suite of engineering design optimization problems and an interface to popular synthetic optimization problems, with access to 300+ problems for optimization algorithm benchmarking.
Our goal is to provide not only a Python optimization benchmark library but also to allow the PyTorch interface for facilitating machine learning optimization algorithms and applications such as surrogate and Bayesian optimization.
> [!IMPORTANT]
>
> The optimization tasks in this library can be used for all kinds of optimization algorithms benchmarking.
> As it was originally designed for Bayesian Optimization algorithms, the output objective values are meant to be **_maximized_**.
> If you are using minimization algorithms, please negate the output objective value for your use.
# What is in BOCoDe?
## Engineering Design Problems
We present a diverse collection of engineering design problems including car design, cantilever beam, truss structure optimization, and physics simulation of robotics problems.
<center><img src="docs/Figures/TopFuns_Icon_v2.png" width="500"></center>
## Interface to Popular Benchmarks
| [Botorch](https://botorch.org/) | [BBOB/COCO](https://coco-platform.org/) | [OPFUNU<br/>(IEEE CEC benchmarks)](https://github.com/thieu1995/opfunu) | [Gym Mujoco](https://www.gymlibrary.dev/environments/mujoco/index.html) | [NEORL](https://neorl.readthedocs.io/en/latest/#) |
| :------: | :------: | :------: | :------: | :------: |
| <img src="docs/Figures/botorch_icon.png" width="50"> | <img src="docs/Figures/coco-logo.svg" width="50"> | <img src="docs/Figures/opfunu.png" width="100"> | <img src="docs/Figures/gym_logo.png" width="120"> | <img src="docs/Figures/Neorl_logo.png" width="50"> |
Other open-source libraries and benchmarks: [MODAct](https://github.com/epfl-lamd/modact), [Lassobench](https://github.com/ksehic/LassoBench), [BayesianCHT](https://github.com/TsaiYK/BayesianCHT), [DTLZ](https://www.research-collection.ethz.ch/handle/20.500.11850/145762), [WFG](https://ieeexplore.ieee.org/document/1705400), [ZDT](https://pubmed.ncbi.nlm.nih.gov/10843520/)
# Installation
You can install the core library from PyPI:
```bash
pip install bocode
```
Full Installation (with External Dependencies)
This library also supports benchmarks from `LassoBench` and `modact`, which are not available on PyPI. To use them, you must install them directly from their Git repositories after installing bocode:
```bash
git clone github.com/rosenyu304/BOCoDe/
cd BOCoDe
pip install -e .[full]
```
# Optimization Problem Definition
Here we define all our problems for **maximization** optimization algorithms (for minimization, negate the evaluated value). We define constraints to be **inequality constraints** (i.e. constraint values (gx) <= 0 as feasible).
<center><img src="docs/Figures/opt_definition.png" width="300"></center>
# Example Usage
For details of each problem's usage, please read our docs. Here we provide examples to common usage of this library:
1. Direct evaluation
```python
import bocode
import torch
# Instantiate a benchmark problem
problem = bocode.Engineering.Car()
# Evaluate at a point
x = torch.Tensor([[0.0] * problem.dim])
values, constraints = problem.evaluate(x)
print(f"Is it feasible? {(constraints<=0).all()}")
print(f"Function value at origin: {values[0]}")
```
2. Scaling parameters sampled from unit hypercube (typical Bayesian optimization practice)
```python
import bocode
import torch
# Instantiate a Synthetic benchmark problem
problem = bocode.Synthetics.Ackley()
# Evaluate at a in bounds of [0,1]s
x = torch.rand(5,problem.dim)
print("X in [0,1]s:\n",x,"\n")
# Scale it w.r.t. the problem bounds
x = problem.scale(x)
print("Scaled X in bounds:\n",x)
values, constraints = problem.evaluate(x)
print(f"Is each sample feasible? {(constraints<=0).all(dim=1)}")
print(f"Function value at origin: {values[0]}")
```
3. Example using a scipy minimization for this
```python
import bocode
import numpy as np
import torch
from scipy.optimize import minimize
# Create a benchmark problem
problem = bocode.Synthetics.Michalewicz(dim=2)
problem.visualize_function()
# Get problem bounds
bounds = problem.bounds
# Define objective function for optimizer
def objective(x):
x = torch.Tensor([x])
fx, _ = problem.evaluate(x)
fx = -fx # Negate the objective function for MINIMIZATION
return fx.numpy()[0][0]
# Starting point (2-dimensional)
x0 = np.zeros(2)
# Optimize using SciPy
result = minimize(objective, x0, method='Powell', bounds=bounds)
print(f"Optimal value found: {result.fun}")
print(f"Optimal point found: {result.x}")
print(f"Actual optimal value: {-problem.optimum[0]}")
print(f"Actual optimal point: {problem.x_opt[0]}")
```
4. Synthetic function visualization
```python
import bocode
import torch
# Instantiate a benchmark problem
problem = bocode.Synthetics.Powell()
# Visualize the function
problem.visualize_function()
```
# Development
BOCoDe is an open source project and we welcome contributions! If you want to add a new problem, please reach out to us first to see if it is a good fit for BOCoDe.
# Citing
1. If you use BOCoDe in your research, please cite the following paper:
```bibtex
@misc{yu2025bocode,
author={Rosen Ting-Ying Yu, Advaith Narayanan, Cyril Picard, Faez Ahmed},
title = {{BOCoDe}: Benchmarks for Optimization and Computational Design},
year={2025},
url={https://github.com/rosenyu304/BOCoDe}
}
```
2. If you use the the BOCoDe interfaces to other libraries or open source code functions (ex: BoTorch, BBOB, NEORL, MODAct, LassoBench, ...), please cite them accordingly.
# License
BOCoDe is MIT licensed, as found in [LICENSE](LICENSE)
Raw data
{
"_id": null,
"home_page": null,
"name": "bocode",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.10",
"maintainer_email": null,
"keywords": "engineering design, optimization, benchmark, engineering, bayesian-optimization",
"author": "Rosen Yu, Advaith Narayanan, Cyril Picard, Faez Ahmed",
"author_email": null,
"download_url": "https://files.pythonhosted.org/packages/35/d8/7d3a5eba434924cbe02dc20bd134d51420c7b822b8b8d801228d942cdb50/bocode-0.1.6.dev0.tar.gz",
"platform": null,
"description": "# BOCoDe: Benchmarks for Optimization and Computational Design\n\nWe present BOCoDe, a Python and PyTorch-based library that contains the most comprehensive suite of engineering design optimization problems and an interface to popular synthetic optimization problems, with access to 300+ problems for optimization algorithm benchmarking. \n\nOur goal is to provide not only a Python optimization benchmark library but also to allow the PyTorch interface for facilitating machine learning optimization algorithms and applications such as surrogate and Bayesian optimization.\n\n> [!IMPORTANT]\n>\n> The optimization tasks in this library can be used for all kinds of optimization algorithms benchmarking. \n> As it was originally designed for Bayesian Optimization algorithms, the output objective values are meant to be **_maximized_**.\n> If you are using minimization algorithms, please negate the output objective value for your use.\n\n# What is in BOCoDe?\n\n## Engineering Design Problems\nWe present a diverse collection of engineering design problems including car design, cantilever beam, truss structure optimization, and physics simulation of robotics problems. \n\n<center><img src=\"docs/Figures/TopFuns_Icon_v2.png\" width=\"500\"></center>\n\n\n## Interface to Popular Benchmarks\n\n| [Botorch](https://botorch.org/) | [BBOB/COCO](https://coco-platform.org/) | [OPFUNU<br/>(IEEE CEC benchmarks)](https://github.com/thieu1995/opfunu) | [Gym Mujoco](https://www.gymlibrary.dev/environments/mujoco/index.html) | [NEORL](https://neorl.readthedocs.io/en/latest/#) |\n| :------: | :------: | :------: | :------: | :------: |\n| <img src=\"docs/Figures/botorch_icon.png\" width=\"50\"> | <img src=\"docs/Figures/coco-logo.svg\" width=\"50\"> | <img src=\"docs/Figures/opfunu.png\" width=\"100\"> | <img src=\"docs/Figures/gym_logo.png\" width=\"120\"> | <img src=\"docs/Figures/Neorl_logo.png\" width=\"50\"> |\n\nOther open-source libraries and benchmarks: [MODAct](https://github.com/epfl-lamd/modact), [Lassobench](https://github.com/ksehic/LassoBench), [BayesianCHT](https://github.com/TsaiYK/BayesianCHT), [DTLZ](https://www.research-collection.ethz.ch/handle/20.500.11850/145762), [WFG](https://ieeexplore.ieee.org/document/1705400), [ZDT](https://pubmed.ncbi.nlm.nih.gov/10843520/)\n\n# Installation\n\nYou can install the core library from PyPI:\n\n```bash\npip install bocode\n```\n\nFull Installation (with External Dependencies)\nThis library also supports benchmarks from `LassoBench` and `modact`, which are not available on PyPI. To use them, you must install them directly from their Git repositories after installing bocode:\n```bash\ngit clone github.com/rosenyu304/BOCoDe/\ncd BOCoDe\npip install -e .[full]\n```\n\n# Optimization Problem Definition\nHere we define all our problems for **maximization** optimization algorithms (for minimization, negate the evaluated value). We define constraints to be **inequality constraints** (i.e. constraint values (gx) <= 0 as feasible).\n\n<center><img src=\"docs/Figures/opt_definition.png\" width=\"300\"></center>\n\n# Example Usage\n\nFor details of each problem's usage, please read our docs. Here we provide examples to common usage of this library:\n\n1. Direct evaluation\n```python\nimport bocode\nimport torch\n\n# Instantiate a benchmark problem\nproblem = bocode.Engineering.Car()\n\n# Evaluate at a point\nx = torch.Tensor([[0.0] * problem.dim])\nvalues, constraints = problem.evaluate(x)\n\nprint(f\"Is it feasible? {(constraints<=0).all()}\")\nprint(f\"Function value at origin: {values[0]}\")\n```\n\n2. Scaling parameters sampled from unit hypercube (typical Bayesian optimization practice)\n```python\nimport bocode\nimport torch\n\n# Instantiate a Synthetic benchmark problem\nproblem = bocode.Synthetics.Ackley()\n\n# Evaluate at a in bounds of [0,1]s\nx = torch.rand(5,problem.dim)\nprint(\"X in [0,1]s:\\n\",x,\"\\n\")\n\n# Scale it w.r.t. the problem bounds\nx = problem.scale(x)\nprint(\"Scaled X in bounds:\\n\",x)\nvalues, constraints = problem.evaluate(x)\n\nprint(f\"Is each sample feasible? {(constraints<=0).all(dim=1)}\")\nprint(f\"Function value at origin: {values[0]}\")\n```\n\n3. Example using a scipy minimization for this\n```python\nimport bocode\nimport numpy as np\nimport torch\nfrom scipy.optimize import minimize\n\n# Create a benchmark problem\nproblem = bocode.Synthetics.Michalewicz(dim=2)\n\nproblem.visualize_function()\n\n# Get problem bounds\nbounds = problem.bounds\n\n# Define objective function for optimizer\ndef objective(x):\n x = torch.Tensor([x])\n fx, _ = problem.evaluate(x)\n fx = -fx # Negate the objective function for MINIMIZATION\n return fx.numpy()[0][0]\n\n# Starting point (2-dimensional)\nx0 = np.zeros(2)\n\n# Optimize using SciPy\nresult = minimize(objective, x0, method='Powell', bounds=bounds)\n\nprint(f\"Optimal value found: {result.fun}\")\nprint(f\"Optimal point found: {result.x}\")\n\nprint(f\"Actual optimal value: {-problem.optimum[0]}\")\nprint(f\"Actual optimal point: {problem.x_opt[0]}\")\n```\n\n4. Synthetic function visualization\n```python\nimport bocode\nimport torch\n\n# Instantiate a benchmark problem\nproblem = bocode.Synthetics.Powell() \n\n# Visualize the function\nproblem.visualize_function()\n```\n\n# Development\n\nBOCoDe is an open source project and we welcome contributions! If you want to add a new problem, please reach out to us first to see if it is a good fit for BOCoDe.\n\n# Citing\n\n1. If you use BOCoDe in your research, please cite the following paper:\n```bibtex\n@misc{yu2025bocode,\n author={Rosen Ting-Ying Yu, Advaith Narayanan, Cyril Picard, Faez Ahmed},\n title = {{BOCoDe}: Benchmarks for Optimization and Computational Design},\n year={2025},\n url={https://github.com/rosenyu304/BOCoDe}\n}\n```\n\n2. If you use the the BOCoDe interfaces to other libraries or open source code functions (ex: BoTorch, BBOB, NEORL, MODAct, LassoBench, ...), please cite them accordingly.\n\n\n# License\nBOCoDe is MIT licensed, as found in [LICENSE](LICENSE)\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "BOCoDe is a Python library contains optimization benchmark problems.",
"version": "0.1.6.dev0",
"project_urls": {
"Homepage": "https://github.com/rosenyu304/BOCoDe/",
"Repository": "https://github.com/rosenyu304/BOCoDe/"
},
"split_keywords": [
"engineering design",
" optimization",
" benchmark",
" engineering",
" bayesian-optimization"
],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "c1c4035527f8288cddb368141af070dacd02875e4302d55e4fd54d768248dbc1",
"md5": "11f66430651512b24b453950a3b64a9b",
"sha256": "895eb9d8206eb9b93d24f1290a0099fb8941b0073f2fc96d33aec649293f4fb5"
},
"downloads": -1,
"filename": "bocode-0.1.6.dev0-py3-none-any.whl",
"has_sig": false,
"md5_digest": "11f66430651512b24b453950a3b64a9b",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.10",
"size": 95914558,
"upload_time": "2025-08-17T17:47:10",
"upload_time_iso_8601": "2025-08-17T17:47:10.396225Z",
"url": "https://files.pythonhosted.org/packages/c1/c4/035527f8288cddb368141af070dacd02875e4302d55e4fd54d768248dbc1/bocode-0.1.6.dev0-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "35d87d3a5eba434924cbe02dc20bd134d51420c7b822b8b8d801228d942cdb50",
"md5": "81b2939cc8e19fd5457a7468eb5cce09",
"sha256": "49afad8cffe9afe7621135aff984d1579c03852ea9c7919ee3ccb364bb460cfa"
},
"downloads": -1,
"filename": "bocode-0.1.6.dev0.tar.gz",
"has_sig": false,
"md5_digest": "81b2939cc8e19fd5457a7468eb5cce09",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.10",
"size": 94893232,
"upload_time": "2025-08-17T17:47:16",
"upload_time_iso_8601": "2025-08-17T17:47:16.087590Z",
"url": "https://files.pythonhosted.org/packages/35/d8/7d3a5eba434924cbe02dc20bd134d51420c7b822b8b8d801228d942cdb50/bocode-0.1.6.dev0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-08-17 17:47:16",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "rosenyu304",
"github_project": "BOCoDe",
"github_not_found": true,
"lcname": "bocode"
}
JSON
