Cognitive gravitation model for classification on small noisy data

Unsupervised band selection of medical hyperspectral images guided by data gravitation and weak correlation

BACKGROUND AND OBJECTIVE

Medical hyperspectral images (MHSIs) are used for a contact-free examination of patients without harmful radiation. However, high-dimensionality images contain large amounts of data that are sparsely distributed in a high-dimensional space, which leads to the "curse of dimensionality" (called Hughes' phenomenon) and increases the complexity and cost of data processing and storage. Hence, there is a need for spectral dimensionality reduction before the clinical application of MHSIs. Some dimensionality-reducing strategies have been proposed; however, they distort the data within MHSIs.

METHODS

To compress dimensionality without destroying the original data structure, we propose a method that involves data gravitation and weak correlation-based ranking (DGWCR) for removing bands of noise from MHSIs while clustering signal-containing bands. Band clustering is done by using the connection centre evolution (CCE) algorithm and selecting the most representative bands in each cluster based on the composite force. The bands within the clusters are ranked using the new entropy-containing matrix, and a global ranking of bands is obtained by applying an S-shaped strategy. The source code is available at https://www.github.com/zhangchenglong1116/DGWCR.

RESULTS

Upon feeding the reduced-dimensional images into various classifiers, the experimental results demonstrated that the small number of bands selected by the proposed DGWCR consistently achieved higher classification accuracy than the original data. Unlike other reference methods (e.g. the latest deep-learning-based strategies), DGWCR chooses the spectral bands with the least redundancy and greatest discrimination.

CONCLUSION

In this study, we present a method for efficient band selection for MHSIs that alleviates the "curse of dimensionality". Experiments were validated with three MHSIs in the human brain, and they outperformed several other band selection methods, demonstrating the clinical potential of DGWCR.

A Geometrical Divide of Data Particle in Gravitational Classification of Moons and Circles Data Sets

Thus far, the Universal Law of Gravitation has found application in many issues related to pattern classification. Its popularity results from its clear theoretical foundations and the competitive effectiveness of the classifiers based on it. Both Moons and Circles data sets constitute distinctive types of data sets that can be found in machine learning. Despite the fact that they have not been formally defined yet, on the basis of their visualization, they can be defined as sets in which the distribution of objects of individual classes creates shapes similar to circles or semicircles. This article makes an attempt to improve the gravitational classifier that creates a data particle based on the class. The aim was to compare the effectiveness of the developed Geometrical Divide method with the popular method of creating a class-based data particle, which is described by a compound of 1 ÷ 1 cardinality in the Moons and Circles data sets classification process. The research made use of eight artificially generated data sets, which contained classes that were explicitly separated from each other as well as data sets with objects of different classes that did overlap each other. Within the limits of the conducted experiments, the Geometrical Divide method was combined with several algorithms for determining the mass of a data particle. The research did also use the k-Fold Cross-Validation. The results clearly showed that the proposed method is an efficient approach in the Moons and Circles data sets classification process. The conclusion section of the article elaborates on the identified advantages and disadvantages of the method as well as the possibilities of further research and development.