2019/07/31 by Henry Wilde, Wilde, Henry, Vincent Knight +3
Computer Science · #Advanced Clustering Algorithms Research #Data Structures and Algorithms (cs.DS) #Evolutionary Algorithms and Applications #FOS: Computer and information sciences #Metaheuristic Optimization Algorithms Research #Neural and Evolutionary Computing (cs.NE)
paper · pdf · doi:10.48550/arxiv.1907.13508
openalex publication_date 2019/07/31 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28
In this paper we propose a novel method for learning how algorithms perform.\nClassically, algorithms are compared on a finite number of existing (or newly\nsimulated) benchmark datasets based on some fixed metrics. The algorithm(s)\nwith the smallest value of this metric are chosen to be the `best performing'.\nWe offer a new approach to flip this paradigm. We instead aim to gain a richer\npicture of the performance of an algorithm by generating artificial data\nthrough genetic evolution, the purpose of which is to create populations of\ndatasets for which a particular algorithm performs well on a given metric.\nThese datasets can be studied so as to learn what attributes lead to a\nparticular progression of a given algorithm. Following a detailed description\nof the algorithm as well as a brief description of an open source\nimplementation, a case study in clustering is presented. This case study\ndemonstrates the performance and nuances of the method which we call\nEvolutionary Dataset Optimisation. In this study, a number of known properties\nabout preferable datasets for the clustering algorithms known as (k)-means and\nDBSCAN are realised in the generated datasets.\n