Welcome to culebra’s documentation!¶
Culebra was born as a DEAP-based evolutionary computation library designed to solve feature selection problems. However, it has been redesigned to support different kind of problems and also different metaheuristics.
Experiments and experiment batchs are automatized by means of the
Experiment and Batch
classes, both in the tools module. Statistical analysis of
the Results is also provided by the
ResultsAnalyzer class.
Culebra is structured in the following modules:
The
abcmodule, which defines the abstract base classes that support culebraThe
checkermodule, which provides several checker functions used within culebra to prevent wrong arguments to functions and methodsThe
fitness_functionmodule, which provides fitness functions for several problemsThe
solutionmodule, which define solutions and solution species for several problemsThe
toolsmodule, which implements several tools to handle data and make easier the experimentation and the analysis of the obtained resultsThe
trainermodule, which implement several training algorithms
Attributes:¶
- DEFAULT_CHECKPOINT_BASENAME = 'checkpoint'¶
Default basename for checkpointing files.
- DEFAULT_INDEX = 0¶
Default index.
- DEFAULT_MAX_NUM_ITERS = 100¶
Default maximum number of iterations.
- DEFAULT_SIMILARITY_THRESHOLD = 0.0¶
Default similarity threshold for fitnesses.
- SERIALIZED_FILE_EXTENSION = '.dill.gz'¶
Extension for files containing serialized objects.
Indices and tables¶
References¶
J. González, J. Ortega, M. Damas, P. Martín-Smith, John Q. Gan. A new multi-objective wrapper method for feature selection - Accuracy and stability analysis for BCI. Neurocomputing, 333:407-418, 2019. https://doi.org/10.1016/j.neucom.2019.01.017.
J. González, J. Ortega, J. J. Escobar, M. Damas. A lexicographic cooperative co-evolutionary approach for feature selection. Neurocomputing, 463:59-76, 2021. https://doi.org/10.1016/j.neucom.2021.08.003.

