🗂️ Versions pulled from the Releases page
v0.2.0 (2025/05/25)
Changes:
- New Algorithm: Implementation of the AIRS algorithm, as detailed in the book Natural Computing Algorithms by Brabazon, O’Neill, and McGarraghy (Springer, 2015).
🔗 View on GitHub
v0.1.42 (2025/05/24)
Changes:
- Standardizes class and method docstrings according to the numpydoc format.
- Adds automated checks to the CI pipeline to ensure ongoing compliance with the docstring standard.
🔗 View on GitHub
v0.1.41 (2025/05/24)
Changes:
- fixes error in examples caused by changing detectors to private field
🔗 View on GitHub
v0.1.40 (2025/05/03)
Changes:
This update aims to improve the package's performance by utilizing Numba for JIT compilation.
-
Python version compatibility:
- Extended support for Python versions 3.10, 3.11, 3.12, and 3.13.
-
New package added: numba:
- Added the dependency
numba >= 0.59.0.
-
Changes to the negative selection algorithm (NSA):
- Renamed the base class to
BaseNSA and introduced new utilities for distance calculation.
- Removed methods such as
_distance, replacing them with utility functions (e.g., compute_metric_distance).
- Refactored the management of detectors (
_detectors).
- Modified the logic of methods such as
fit, predict, and detector validation, utilizing the Numba decorator in critical sections.
-
New module:
-
Added a new base module for methods shared between different classes, such as the BaseClassifier class for classification methods.
🔗 View on GitHub
v0.1.34 (2025/04/06)
Changes:
- Removed redundant
score methods from subclasses of Base. Renamed _score to score in Base, making it public and centralizing the logic.
🔗 View on GitHub
v0.1.33 (2024/10/14)
Changes:
- The package was refactored with the addition of the
utils module, which centralizes common functions for reuse in future modules.
- Unit tests were added for the
utils module.
- In-code comments were updated to English.
🔗 View on GitHub
AISP - 0.1.32 (2024/05/05)
Changes:
- The generation and comparison of detectors using the Hamming distance in the BNSA class has been replaced by the
cdist function from scipy to improve generation and prediction performance.
🔗 View on GitHub
AISP - 0.1.31 (2024/04/29)
Changes:
Addition:
- A new method for selecting labels for samples classified as non-self by all detectors.
- The examples have been adjusted to demonstrate this addition.
Refactoring:
- The source code has been reviewed and adjusted according to the official Python PEP8 guide.
- Repositioning of files for better readability and flexibility.
🔗 View on GitHub
AISP - 0.1.30 (2023/07/01)
Changes:
Bug fix: The class assignment method for samples that trigger all non-self detectors and become non-self for all classes has been fixed. Now, the method uses the class whose detectors have the greatest average distance from the sample.
🔗 View on GitHub
AISP - 0.1.21 (2023/06/15)
Changes:
- Highlighting the addition of an optional variable in the constructor that allows changing the default value of
p==2 in the Minkowski distance.
🔗 View on GitHub
AISP - 0.1.1 (2023/06/10)
Changes:
- Adds the negative selection module to the package.
🔗 View on GitHub