iOpt 0.4.0

Framework for automatically tuning hyperparameter values for mathematical models, AI and ML.

iOpt is an open source framework for automatic selection of parameter values both for mathematical models of complex industrial processes and for AI and ML methods used in industry. The framework is distributed under the 3-Clause BSD license.

Key features of the framework

• Automatic selection of parameter values both for mathematical models and for AI and ML methods used in industry.
• Intelligent control of the process of choosing the optimal parameters for industrial applications.
• Integration with external artificial intelligence and machine learning libraries or frameworks as well as applied models.
• Automation of the preliminary analysis of the models under study, e.g., by identifying different types of model dependencies on different groups of parameters.
• Visualization of the process of choosing optimal parameters.

Installation

Automatic installation

The simplest way to install iOpt is using pip:

pip install iOpt

Manual installation

On Unix-like systems:

git clone https://github.com/aimclub/iOpt
cd iOpt
pip install virtualenv
virtualenv ioptenv
source ioptenv/bin/activate
python setup.py install

On Windows:

git clone https://github.com/aimclub/iOpt
cd iOpt
pip install virtualenv
virtualenv ioptenv
ioptenv\Scripts\activate.bat
python setup.py install

Docker

Download the image:

docker pull aimclub/iopt:latest

Using the iOpt image:

docker run -it aimclub/iopt:latest

How to Use

Using the iOpt framework to minimize the Rastrigin test function.

from problems.rastrigin import Rastrigin
from iOpt.solver import Solver
from iOpt.solver_parametrs import SolverParameters
from iOpt.output_system.listeners.static_painters import StaticPainterNDListener
from iOpt.output_system.listeners.console_outputers import ConsoleOutputListener

from subprocess import Popen, PIPE, STDOUT

if __name__ == "__main__":
    """
    Minimization of the Rastrigin test function with visualization
    """
    # Create a test task
    problem = Rastrigin(2)
    # Setup a solver options
    params = SolverParameters(r=2.5, eps=0.01, iters_limit=300, refine_solution=True)
    # Create the solver
    solver = Solver(problem, parameters=params)
    # Print results to console while solving
    cfol = ConsoleOutputListener(mode='full')
    solver.add_listener(cfol)
    # 3D visualization at the end of the solution
    spl = StaticPainterNDListener("rastrigin.png", "output", vars_indxs=[0, 1], mode="surface", calc="interpolation")
    solver.add_listener(spl)
    # Run problem solution
    sol = solver.solve()

Examples

Let’s demonstrate the use of the iOpt framework when tuning the hyperparameters of one of the machine learning methods. In the support vector machine (SVC), we find the optimal hyperparameters (the regularization parameter C, the kernel coefficient gamma) in the problem of breast cancer classification (detailed description of the data).

import numpy as np
from sklearn.utils import shuffle
from sklearn.datasets import load_breast_cancer

from iOpt.output_system.listeners.static_painters import StaticPainterNDListener
from iOpt.output_system.listeners.animate_painters import AnimatePainterNDListener
from iOpt.output_system.listeners.console_outputers import ConsoleOutputListener
from iOpt.solver import Solver
from iOpt.solver_parametrs import SolverParameters
from examples.Machine_learning.SVC._2D.Problems import SVC_2d


def load_breast_cancer_data():
    dataset = load_breast_cancer()
    x_raw, y_raw = dataset['data'], dataset['target']
    inputs, outputs = shuffle(x_raw, y_raw ^ 1, random_state=42)
    return inputs, outputs


if __name__ == "__main__":
    x, y = load_breast_cancer_data()
    regularization_value_bound = {'low': 1, 'up': 6}
    kernel_coefficient_bound = {'low': -7, 'up': -3}

    problem = SVC_2d.SVC_2D(x, y, regularization_value_bound, kernel_coefficient_bound)

    method_params = SolverParameters(r=np.double(3.0), iters_limit=100)
    solver = Solver(problem, parameters=method_params)

    apl = AnimatePainterNDListener("svc2d_anim.png", "output", vars_indxs=[0, 1], to_paint_obj_func=False)
    solver.add_listener(apl)

    spl = StaticPainterNDListener("svc2d_stat.png", "output", vars_indxs=[0, 1], mode="surface", calc="interpolation")
    solver.add_listener(spl)

    cfol = ConsoleOutputListener(mode='full')
    solver.add_listener(cfol)

    solver_info = solver.solve()

Project Structure

The latest stable release of iOpt is in the main branch. The repository includes the following directories:

• The iOpt directory contains the framework core in the form of Python classes.
• The examples directory contains examples of using the framework for both test and applied problems.
• Unit tests are located in the test directory.
• Documentation source files are located in the docs directory.

Documentation

A detailed description of the iOpt framework API is available at Read the Docs.

Supported by

The study is supported by the Research Center Strong Artificial Intelligence in Industry of ITMO University as part of the plan of the center's program: Framework of intelligent heuristic optimization methods.