Special Issue on “New Advances in Similarity-Based Systems and Applications”

Similarity-based systems are considered to be of central interest for artificial intelligence and cognitive sciences research (Hahn and Chater 1998; Sloman and Rips 1998). Similarity-based systems describe analogies in human reasoning (Pothos 2005). The basic idea behind similarity-based systems is that, in general, reasoning concerned with either comparing different situations or analyzing new situations depends on the similarity between one or more past situations (Hahn and Chater 1998; Stanfill and Waltz 1986).

The main theoretical research approaches on similarity-based systems are founded on mathematical concepts; basically on distance or metrics studies and topologies. However, applications of similarity-based systems are focused mainly on the area of classification and learning, ranging from case-based reasoning studies and implicit learning to online recommender systems (Burkhard 2001; Bridge and Ferguson 2002).

There are several challenges that could be highlighted in this area of research: from a theoretical point of view, challenges related to the use of nonmonotonic logics to find new approaches that allow us to capture the uncertainty defined in human reasoning and decision-making; from a technical point of view, challenges to achieve scalable similarity–based systems able to operate on Big Data, and the necessity of obtaining similarity-based systems for privacy-preserving searches (Zezula 2014). And finally, challenges are considered in application areas such as bioinformatics, engineering systems, and social sciences.

The purpose of the present Special Issue of Applied Artificial Intelligence: An International Journal is to communicate and present some of the latest research carried out in this area, which was presented at the 17th International Conference of the Catalan Association on Artificial Intelligence, held in Barcelona in October 2014. This Special Issue includes extended versions of the best six contributions selected in the field of new advances in similarity-based systems and applications.

The first article, “A Logical Study of Local and Global Graded Similarities,” examines the relationship between global and local similarities in the graded framework of fuzzy class theory, in which a graded notion of similarity already exists.

Then, a comparative experimental study between similarity learning and feature-based and distance-based representations for computing similarity scores is presented in the article “Classification Similarity Learning Using Feature-Based and Distance-Based Representations: A Comparative Study.” The article uses the Support Vector Machine paradigm as a flexible combiner both for a high-dimensional feature space and for a family of distance measures in order to finally learn similarity scores. The approaches are tested in a content-based image retrieval context, using three different repositories.

In the third article, “A Randomized Algorithm for the Exact Solution of Transductive Support Vector Machines,” a method to approximate the value function able to generalize in continuous (or large) space reinforcement learning (RL) problems, is presented.

“A Web-Based Environment to Support Online and Collaborative Group Recommendation Scenarios” presents and demonstrates, through a live-user case-study, the usability of a novel web-based environment that supports online group recommendation. Users were requested to join in a group discussion by using an initial webpage of the interface and then performing a search task of their favorite ski vacation.

A method for extracting the vector of preferences for a set-valued preference matrix and its relation to hesitant fuzzy preference relations is presented in the fifth article, “Derivation of Priorities and Weights for Set-Valued Matrices Using the Geometric Mean Approach.”

The final article, “Generation and Characterization of Fuzzy T-preorders,” studies how, from a given t-norm, T-preorders can be generated in a natural way by a single fuzzy subset. This is considered as an approach for similarity-based generalization of fuzzy orderings. T-preorders and strong complete T-preorders, as well as different fuzzy subsets generating the same T-preorder are characterized.