Stud Hybrid Tasmanian Devil - Grey Wolf Optimization: A novel Bio-Inspired Optimization Algorithm for Learning-to-Rank

Abstract

<p>Building ranking models used in Information Retrieval (IR) is one of the main problems of learning-to-Rank. Tasmanian Devil Optimization algorithm is a recently proposed meta-heuristic optimization algorithm based on the Tasmanian Devils’ predation mechanism. In this paper, grey wolf operator and stud mechanism are introduced into the TDO optimization process to improve the ability of exploitation and achieve better performance. The proposed Stud Hybrid Tasmanian Devil - Grey Wolf Optimization (SHTDO) combines Grey Wolf Optimizer and TDO with a Stud Selection and Crossover Operator, thus helping to enhance the exploration and exploitation as well as the motion under both strategies. Also, to validate the algorithm on the different fields, widely accepted benchmark and dataset was used for testing. The results show that our algorithm obtains higher ranking performance than TDO, which makes it more suitable for solving optimization problems, especially LTR tasks.</p> <p> </p>