China
Africa
Explore chapters and articles related to this topic
Leveraging Center-Based Sampling Theory for Enhancing Particle Swarm Classification of Textual Data
Published in Wellington Pinheiro dos Santos, Juliana Carneiro Gomes, Valter Augusto de Freitas Barbosa, Swarm Intelligence Trends and Applications, 2023
CBS-PSO is a PSO variant developed specifically to tackle data classification tasks. It is inspired by center-based sampling theory, which emphasizes the usefulness of the center region of the search space for population-based algorithms, particularly when applied to high dimensional spaces (Rahnamayan and Wong, 2009; Esmailzadeh and Rahnamayan, 2011). According to this theory, as a point they move closer to the/center of the search space, its average distance to the optimal solution becomes lower, and for higher dimensions this distance decreases even sharply. In light of this, the CBS-PSO hypothesizes that attracting the search toward the center region of the search space provides a higher chance of convergence to the optimal solution. To realize this hypothesis, it is required to develop mechanisms to identify the center point of the search space and guide the search toward the center region of the search space. In doing so, the CBS-P SO uses Rocchio algorithm (Rocchio, 1971), an efficient information retrieval algorithm, as an estimation method of the center point of the search for the data classification problem. RA is used as a centroid-based classifier to generate for each class c a prototype vector, which is the average vector overall training set vectors that belong to the class c and uses it to classify a new data instance by calculating the similarity between the vector of the new data instance and each of prototype vectors and assigns it to the class with maximum similarity. The main advantage of the RA method is its simplicity and efficiency in terms of computation time, which is linear in the dataset size and the number of classes (Aggarwal and Zhai, 2012). Moreover, the CBS-PSO uses the generated RA estimation of the center point to generate informed particles and incorporate them in the swarm to attract the PSO evolution toward the center region of the search space. The rationale behind this lies in the searching behavior of the swarm, in which the velocities of particles are determined by their previous velocities, cognitive learning, and social learning. While the social learning drives all particles to be attracted by the global best particle and move toward it, the other two parts, previous velocities and cognitive learning, are corresponded to the autonomy property, which makes particles keep their own information. Therefore, during the search all particles move in the region where the global best is located, but their positions are usually different and approximately around the global best (Liu et al., 2007). Following this, the incorporation of the RA-based informed particle in the CBS-PSO swarm, considering that its high chance to be selected as the global best, it will definitely attracts all particles toward the center region of the search space where it is located, and ultimately enables the swarm to converge around this promising region.
Block-based pseudo-relevance feedback for image retrieval
Published in Journal of Experimental & Theoretical Artificial Intelligence, 2022
The Rocchio algorithm can be regarded as a vector modification approach that bases its iterative reformulation of the query vector on the feedback set, again moving the query towards a topological region of more relevant images and away from irrelevant ones (Rocchio, 1971). It reformulates the query as the modified query by