opensbt.algorithm package

Subpackages

Submodules

opensbt.algorithm.algorithm module

class opensbt.algorithm.algorithm.AlgorithmType(value)[source]

Bases: Enum

Defines all available algorithms to be triggered via number in -a flag.

NSGAII = 1
NSGAII_DT = 6
PSO = 2
PS_FPS = 5
PS_GRID = 4
PS_RAND = 3

opensbt.algorithm.nsga2_optimizer module

class opensbt.algorithm.nsga2_optimizer.NsgaIIOptimizer(problem: Problem, config: SearchConfiguration)[source]

Bases: Optimizer

This optimizer class provides the NSGA-II algorithm which is already implemented in pymoo.

__init__(problem: Problem, config: SearchConfiguration)[source]

Initialize here the Optimization algorithm to be used for search-based testing.

Parameters:
_abc_impl = <_abc._abc_data object>
algorithm: Algorithm

Prioritize max search time over set maximal number of generations

algorithm_name: str = 'NSGA-II'

opensbt.algorithm.nsga2dt_optimizer module

class opensbt.algorithm.nsga2dt_optimizer.NsgaIIDTOptimizer(problem: Problem, config: SearchConfiguration)[source]

Bases: Optimizer

This optimizer implements the NSGA-II-DT algorithm from [1] which is based on the NSGA-II algorithm but is employing ML models, i.e. decision trees to guide the search for failures.

[1] Raja Ben Abdessalem, Shiva Nejati, Lionel C. Briand, and Thomas Stifter. 2018. Testing vision-based control systems using learnable evolutionary algorithms. In Proceedings of the 40th International Conference on Software Engineering (ICSE ‘18). Association for Computing Machinery, New York, NY, USA, 1016–1026. https://doi.org/10.1145/3180155.3180160

__init__(problem: Problem, config: SearchConfiguration)[source]

Initializes the NSGA-II-DT Optimizer

Parameters:

  • problem (Problem) – The testing problem.

  • config (SearchConfiguration) – The configuraiton for the search.

_abc_impl = <_abc._abc_data object>
_create_result(problem, hist_holder, inner_algorithm, execution_time)[source]
_update_history(res, hist_holder, tree_iteration, inner_num_gen, inner_algorithm)[source]
algorithm_name: str = 'NSGA-II-DT'
run() → SimulationResult[source]

_summary_

Returns:

Return a SimulationResults object which holds all information from the simulation.

Return type:

SimulationResult

opensbt.algorithm.optimizer module

class opensbt.algorithm.optimizer.Optimizer(problem: SimulationProblem, config: SearchConfiguration)[source]

Bases: ABC

Base class for all optimizers in OpenSBT. Subclasses need to implement the __init__ method. The run method has to be overriden when non pymoo implemented algorithms are used. For reference consider the implementation of the NSGA-II-DT optimizer in opensbt/algorithm/nsga2dt_optimizer.py

abstract __init__(problem: SimulationProblem, config: SearchConfiguration)[source]

Initialize here the Optimization algorithm to be used for search-based testing.

Parameters:
_abc_impl = <_abc._abc_data object>
algorithm: Algorithm
algorithm_name: str
config: SearchConfiguration
parameters: Dict
problem: Problem
run() → SimulationResult[source]

Runs the optimizer for a given problem and search configuration. Returns the simulation output as an instance of SimulationResult. This methods need to overriden when a non pymoo-based algorithm is used (e.g., NSGA-II-DT)

save_history: bool
termination: object

opensbt.algorithm.ps module

class opensbt.algorithm.ps.PureSampling(problem: ~opensbt.model_ga.problem.SimulationProblem, config: ~opensbt.experiment.search_configuration.SearchConfiguration, sampling_type=<class 'pymoo.operators.sampling.rnd.FloatRandomSampling'>)[source]

Bases: Optimizer

This class provides the parent class for all sampling based search algorithms.

__init__(problem: ~opensbt.model_ga.problem.SimulationProblem, config: ~opensbt.experiment.search_configuration.SearchConfiguration, sampling_type=<class 'pymoo.operators.sampling.rnd.FloatRandomSampling'>)[source]

Initializes pure sampling approaches.

Parameters:

  • problem (Problem) – The testing problem to be solved.

  • config (SearchConfiguration) – The configuration for the search.

  • sampling_type (_type_, optional) – Sets by default sampling type to RS.

_abc_impl = <_abc._abc_data object>
_create_result(problem, pop, execution_time, n_splits)[source]
algorithm_name: str = 'RS'
run() → SimulationResult[source]
Overrides the run method of Optimizer by providing custom evaluation of samples and division in “buckets” for further analysis with pymoo.

(s. n_splits variable)

Returns:

Return a SimulationResults object which holds all information from the simulation.

Return type:

SimulationResult

opensbt.algorithm.ps_fps module

class opensbt.algorithm.ps_fps.PureSamplingFPS(problem: ~opensbt.model_ga.problem.SimulationProblem, config: ~opensbt.experiment.search_configuration.SearchConfiguration, sampling_type=<class 'opensbt.utils.fps.FPS'>)[source]

Bases: PureSampling

This class provides the Farthest Point Sampling algorithm (FPS) [1]. FPS generate new inputs with the largest distance from the existing inputs to achieve a uniform distribution of test inputs in the search space.

[1] Eldar Y, Lindenbaum M, Porat M, Zeevi Y (1997) The farthest point strategy for progressive image sampling. IEEE Trans Image Process 6:1305–15, DOI 10.1109/83.623193

__init__(problem: ~opensbt.model_ga.problem.SimulationProblem, config: ~opensbt.experiment.search_configuration.SearchConfiguration, sampling_type=<class 'opensbt.utils.fps.FPS'>)[source]

Initializes the furthes point sampling optimizer.

Parameters:

  • problem (Problem) – The testing problem to be solved.

  • config (SearchConfiguration) – The configuration for the search.

  • sampling_type (_type_, optional) – Sets by default sampling type to FPS.

_abc_impl = <_abc._abc_data object>
algorithm_name: str = 'FPS'

opensbt.algorithm.ps_grid module

class opensbt.algorithm.ps_grid.PureSamplingGrid(problem: ~opensbt.model_ga.problem.SimulationProblem, config: ~opensbt.experiment.search_configuration.SearchConfiguration, sampling_type=<class 'opensbt.utils.sampling.CartesianSampling'>)[source]

Bases: PureSampling

This class provides the Grid Sampling algorithm which generate aquidistant test inputs placed on a grid in the search space.

__init__(problem: ~opensbt.model_ga.problem.SimulationProblem, config: ~opensbt.experiment.search_configuration.SearchConfiguration, sampling_type=<class 'opensbt.utils.sampling.CartesianSampling'>)[source]

Initializes the grid search sampling optimizer.

Parameters:

  • problem (Problem) – The testing problem to be solved.

  • config (SearchConfiguration) – The configuration for the search. The number of samples is equaly for each axis. The axis based sampling number is defined via the population size.

  • sampling_type (_type_, optional) – Sets by default sampling type to Cartesian Sampling.

_abc_impl = <_abc._abc_data object>

opensbt.algorithm.ps_rand module

class opensbt.algorithm.ps_rand.PureSamplingRand(problem: ~opensbt.model_ga.problem.SimulationProblem, config: ~opensbt.experiment.search_configuration.SearchConfiguration, sampling_type=<class 'pymoo.operators.sampling.rnd.FloatRandomSampling'>)[source]

Bases: PureSampling

This class provides the random sampling algorithm which generates random test inputs in the search space.

__init__(problem: ~opensbt.model_ga.problem.SimulationProblem, config: ~opensbt.experiment.search_configuration.SearchConfiguration, sampling_type=<class 'pymoo.operators.sampling.rnd.FloatRandomSampling'>)[source]

Initializes the random sampling optimizer.

Parameters:

  • problem (Problem) – The testing problem to be solved.

  • config (SearchConfiguration) – The configuration for the search.

  • sampling_type (_type_, optional) – Sets by default sampling type to RS.

_abc_impl = <_abc._abc_data object>

opensbt.algorithm.pso_optimizer module

class opensbt.algorithm.pso_optimizer.PSOOptimizer(problem: SimulationProblem, config: SearchConfiguration)[source]

Bases: Optimizer

This class provides search with the Particle Swarm Optimization algorithm which is already implemented in pymoo.

__init__(problem: SimulationProblem, config: SearchConfiguration)[source]

Initializes the particle swarm opimization approach for testing.

Parameters:
_abc_impl = <_abc._abc_data object>
algorithm: Algorithm

Prioritize max search time over set maximal number of generations

algorithm_name: str = 'PSO'

Module contents