# Evolutune
A Genetic Algorithm-based hyperparameter tuner for machine learning models.
## Introduction
Evolutune, implements a hyperparameter tuner based on the principles of a genetic algorithm. The genetic algorithm evolves a population of hyperparameter sets over several generations, aiming to find the set that optimizes a given scoring metric. This tuner is designed to work with various machine learning models.
## Dependencies
Make sure you have the following dependencies installed:
- ```numpy```
- ```joblib.Parallel``` and ```joblib.delayed```
- ```sklearn.metrics.get_scorer```
## Installation
```sh
pip install evolutune
```
## Usage
```python3
from evolutune import GeneticTuner
# Define your machine learning model
# model = ...
# Define the hyperparameter search space
param_grid = {
'param1': [value1, value2, ...],
'param2': [value3, value4, ...],
# Add more hyperparameters as needed
}
# Define the scoring metric to optimize
scoring_metric = 'accuracy' # Replace with your preferred metric
# Instantiate the GeneticTuner
genetic_tuner = GeneticTuner(
model=model,
param_grid=param_grid,
scoring=scoring_metric,
population_size=10,
generations=100,
mutation_rate=0.1,
random_state=None,
n_jobs=None
)
```
## Fitting the Tuner
```python3
# Define your training and evaluation sets
train_set = [X_train, y_train]
eval_set = [X_eval, y_eval] # Set to None to use the training set for evaluation
# Specify the optimization direction ('maximize' or 'minimize')
direction = 'maximize'
# Fit the tuner on the training set
genetic_tuner.fit(train_set, eval_set, direction)
```
## Accessing Results
```python3
# Access the best score and corresponding hyperparameters
best_score = genetic_tuner.best_score_
best_params = genetic_tuner.best_params_
print(f"Best Score: {best_score}")
print("Best Hyperparameters:")
for param, value in best_params.items():
print(f"{param}: {value}")
```
## Methods
| Method | Description |
|---------------------------------------------------------------------------------|-----------------------------------------------------------------------|
| `initialize_population(population_size: int) -> list` | Initialize a population of individuals with random hyperparameters. |
| `crossover(parent1: dict, parent2: dict) -> tuple` | Perform crossover between two parents to generate two children. |
| `mutate(individual: dict, mutation_rate: float) -> dict` | Introduce random mutations to an individual's hyperparameters. |
| `calculate_fitness(train_set: list, eval_set: list, parameters: dict) -> float` | Evaluate the fitness (scoring metric) of a set of hyperparameters. |
| `fit(train_set: list, eval_set: list = None, direction: str = "maximize")` | Fit the GeneticTuner on the training set and optional evaluation set. |
## Example
An example script demonstrating the usage of the GeneticTuner class is provided in the example.py file.
Raw data
{
"_id": null,
"home_page": "",
"name": "evolutune",
"maintainer": "",
"docs_url": null,
"requires_python": "",
"maintainer_email": "",
"keywords": "python,hyperparameter,tuning,genetic-algorithm,model,search,CV",
"author": "torchd3v",
"author_email": "<burak96egeli@gmail.com>",
"download_url": "https://files.pythonhosted.org/packages/06/31/7b44ef24b2e12a34d120048850068145b672c04812f985311ea8e5591392/evolutune-0.0.3.tar.gz",
"platform": null,
"description": "# Evolutune\nA Genetic Algorithm-based hyperparameter tuner for machine learning models.\n\n## Introduction\nEvolutune, implements a hyperparameter tuner based on the principles of a genetic algorithm. The genetic algorithm evolves a population of hyperparameter sets over several generations, aiming to find the set that optimizes a given scoring metric. This tuner is designed to work with various machine learning models.\n\n## Dependencies\nMake sure you have the following dependencies installed:\n\n- ```numpy```\n- ```joblib.Parallel``` and ```joblib.delayed```\n- ```sklearn.metrics.get_scorer```\n\n## Installation\n```sh\npip install evolutune\n```\n\n## Usage\n```python3\nfrom evolutune import GeneticTuner\n\n# Define your machine learning model\n# model = ...\n\n# Define the hyperparameter search space\nparam_grid = {\n 'param1': [value1, value2, ...],\n 'param2': [value3, value4, ...],\n # Add more hyperparameters as needed\n}\n\n# Define the scoring metric to optimize\nscoring_metric = 'accuracy' # Replace with your preferred metric\n\n# Instantiate the GeneticTuner\ngenetic_tuner = GeneticTuner(\n model=model,\n param_grid=param_grid,\n scoring=scoring_metric,\n population_size=10,\n generations=100,\n mutation_rate=0.1,\n random_state=None,\n n_jobs=None\n)\n```\n\n## Fitting the Tuner\n```python3\n# Define your training and evaluation sets\ntrain_set = [X_train, y_train]\neval_set = [X_eval, y_eval] # Set to None to use the training set for evaluation\n\n# Specify the optimization direction ('maximize' or 'minimize')\ndirection = 'maximize'\n\n# Fit the tuner on the training set\ngenetic_tuner.fit(train_set, eval_set, direction)\n```\n\n## Accessing Results\n```python3\n# Access the best score and corresponding hyperparameters\nbest_score = genetic_tuner.best_score_\nbest_params = genetic_tuner.best_params_\n\nprint(f\"Best Score: {best_score}\")\nprint(\"Best Hyperparameters:\")\nfor param, value in best_params.items():\n print(f\"{param}: {value}\")\n```\n\n## Methods\n\n| Method | Description |\n|---------------------------------------------------------------------------------|-----------------------------------------------------------------------|\n| `initialize_population(population_size: int) -> list` | Initialize a population of individuals with random hyperparameters. |\n| `crossover(parent1: dict, parent2: dict) -> tuple` | Perform crossover between two parents to generate two children. |\n| `mutate(individual: dict, mutation_rate: float) -> dict` | Introduce random mutations to an individual's hyperparameters. |\n| `calculate_fitness(train_set: list, eval_set: list, parameters: dict) -> float` | Evaluate the fitness (scoring metric) of a set of hyperparameters. |\n| `fit(train_set: list, eval_set: list = None, direction: str = \"maximize\")` | Fit the GeneticTuner on the training set and optional evaluation set. |\n\n\n## Example\nAn example script demonstrating the usage of the GeneticTuner class is provided in the example.py file.\n",
"bugtrack_url": null,
"license": "",
"summary": "A Genetic Algorithm-based hyperparameter tuner for machine learning models.",
"version": "0.0.3",
"project_urls": null,
"split_keywords": [
"python",
"hyperparameter",
"tuning",
"genetic-algorithm",
"model",
"search",
"cv"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "8510ecb229951c647789580d777e413d0a42d6e68872d92d65c22c5adf4bfb0e",
"md5": "0b12ab13bfc64f0f3f314015f0450878",
"sha256": "0e625cc7a46c9cd5761297ba5a2fef6bdb59a47fd3ebf1c1af114dd208e37d9d"
},
"downloads": -1,
"filename": "evolutune-0.0.3-py3-none-any.whl",
"has_sig": false,
"md5_digest": "0b12ab13bfc64f0f3f314015f0450878",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": null,
"size": 9289,
"upload_time": "2024-03-05T16:19:28",
"upload_time_iso_8601": "2024-03-05T16:19:28.971467Z",
"url": "https://files.pythonhosted.org/packages/85/10/ecb229951c647789580d777e413d0a42d6e68872d92d65c22c5adf4bfb0e/evolutune-0.0.3-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "06317b44ef24b2e12a34d120048850068145b672c04812f985311ea8e5591392",
"md5": "1f2c1628faf7f5ee3723b3b76095d71f",
"sha256": "a7c782e20709b1d3419a536e5939ba6bff43c27438abd06ed273312cf22ce780"
},
"downloads": -1,
"filename": "evolutune-0.0.3.tar.gz",
"has_sig": false,
"md5_digest": "1f2c1628faf7f5ee3723b3b76095d71f",
"packagetype": "sdist",
"python_version": "source",
"requires_python": null,
"size": 8713,
"upload_time": "2024-03-05T16:19:29",
"upload_time_iso_8601": "2024-03-05T16:19:29.933783Z",
"url": "https://files.pythonhosted.org/packages/06/31/7b44ef24b2e12a34d120048850068145b672c04812f985311ea8e5591392/evolutune-0.0.3.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-03-05 16:19:29",
"github": false,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"lcname": "evolutune"
}
JSON