RNSA

Real-Valued Negative Selection Algorithm (RNSA).

Inheritance

This class extends BaseClassifier

Module: aisp.nsa
Import: from aisp.nsa import RNSA

---

Overview

Algorithm for classification and anomaly detection Based on self or not self discrimination, inspired by Negative Selection Algorithm.

note

This algorithm has two different versions: one based on the canonical version ¹ and another with variable radius detectors ². Both are adapted to work with multiple classes and have methods for predicting data present in the non-self region of all detectors and classes.

warning

The parameters r and r_s can prevent the generation of valid detectors. A very small r value can limit coverage, while a very high one can hinder the generation of valid detectors. Similarly, a high r_s can restrict detector creation. Thus, proper adjustment of r and r_s is essential to ensure good model performance.

---

Example

import numpy as np
from aisp.nsa import RNSA

np.random.seed(1)
class_a = np.random.uniform(high=0.5, size=(50, 2))
class_b = np.random.uniform(low=0.51, size=(50, 2))

Example 1: Multiclass classification (RNSA supports two or more classes)

x_train = np.vstack((class_a, class_b))
y_train = ['a'] * 50 + ['b'] * 50
rnsa = RNSA(N=150, r=0.3, seed=1)
rnsa = rnsa.fit(x_train, y_train, verbose=False)
x_test = [
    [0.15, 0.45],  # Expected: Class 'a'
    [0.85, 0.65],  # Expected: Class 'b'
]
y_pred = rnsa.predict(x_test)
print(y_pred)

Output

['a' 'b']

Example 2: Anomaly Detection (self/non-self)

rnsa = RNSA(N=150, r=0.3, seed=1)
rnsa = rnsa.fit(X=class_a, y=np.array(['self'] * 50), verbose=False)
y_pred = rnsa.predict(class_b[:5])
print(y_pred)

Output

['non-self' 'non-self' 'non-self' 'non-self' 'non-self']

---

Constructor Parameters

Name	Type	Default	Description
N	int	100	Number of detectors.
r	float	0.05	Radius of the detector.
r_s	float	0.0001	rₛ Radius of the X own samples.
k	int	1	Number of neighbors near the randomly generated detectors to perform the distance average calculation.
metric	{"euclidean", "minkowski", "manhattan"}	'euclidean'	Distance metric used to compute the distance between the detector and the sample.
max_discards	int	1000	This parameter indicates the maximum number of consecutive detector discards, aimed at preventing a possible infinite loop in case a radius is defined that cannot generate non-self detectors.
seed	Optional[int]	None	Seed for the random generation of values in the detectors.
algorithm	{"default-NSA", "V-detector"}	'default-NSA'	Set the algorithm version
non_self_label	str	'non-self'	This variable stores the label that will be assigned when the data has only one output class, and the sample is classified as not belonging to that class.
cell_bounds	bool	False	If set to True, this option limits the generation of detectors to the space within the plane between 0 and 1. This means that any detector whose radius exceeds this limit is discarded, this variable is only used in the V-detector algorithm.
p	float	2.0	This parameter stores the value of p used in the Minkowski distance.

Attributes

Name	Type	Default	Description
detectors	Optional[Dict[str | int, list[Detector]]]	-	The trained detectors, organized by class.

---

Public Methods

fit

def fit(
    self,
    X: Union[npt.NDArray, list],
    y: Union[npt.NDArray, list],
    verbose: bool = True,
) -> BNSA:
    ...

Perform training according to X and y, using the negative selection method (NegativeSelect).

Parameters

Name	Type	Default	Description
X	Union[npt.NDArray, list]	-	Training input samples. Each row corresponds to a samples and column to feature.
y	Union[npt.NDArray, list]	-	Target vector of shape (n_samples,). Must contain the same number of samples as X.
verbose	bool	True	If True, prints training progress information.

Returns

Type	Description
Self	Returns the instance itself.

Raises

Exception	Description
TypeError	If X or y are not ndarrays or have incompatible shapes.
ValueError	If the array X fall outside the interval (0.0, 1.0).
MaxDiscardsReachedError	The maximum number of detector discards was reached during maturation. Check the defined radius value and consider reducing it.

---

predict

def predict(self, X: Union[npt.NDArray, list]) -> npt.NDArray:
    ...

Prediction of classes based on detectors created after training.

Parameters

Name	Type	Default	Description
X	Union[npt.NDArray, list]	-	Input samples. Must have the same number of features used during training.

Returns

Type	Description
npt.NDArray	An ndarray of the form C (n_samples), containing the predicted classes for X.

Raises

Exception	Description
TypeError	If X is not a ndarray or list.
ValueError	If the array X fall outside the interval (0.0, 1.0).
FeatureDimensionMismatch	If the number of features in X does not match the expected number.
ModelNotFittedError	If the mode has not yet been adjusted and does not have defined detectors or classes, it is not able to predictions

---

Extended Example

Complete usage examples are available in the Jupyter Notebooks:

• Examples

---

References

Footnotes

1. BRABAZON, Anthony; O'NEILL, Michael; MCGARRAGHY, Seán. Natural Computing Algorithms. [S. l.]: Springer Berlin Heidelberg, 2015. DOI 10.1007/978-3-662-43631-8. Disponível em: https://dx.doi.org/10.1007/978-3-662-43631-8. ↩

2. Ji, Z.; Dasgupta, D. (2004). Real-Valued Negative Selection Algorithm with Variable-Sized Detectors. In Lecture Notes in Computer Science, vol. 3025. https://doi.org/10.1007/978-3-540-24854-5_30 ↩