We investigate classifiers in the sample compression framework that can be specified by two distinct sources of information: a compression set and a message string of additional information. In the compression setting, a reconstruction function specifies a classifier when given this information. We examine how an efficient redistribution of this reconstruction information can lead to more general classifiers. In particular, we derive risk bounds that can provide an explicit control over the sparsity of the classifier and the magnitude of its separating margin and a capability to perform a margin-sparsity trade-off in favor of better classifiers. We show how an application to the set covering machine algorithm results in novel learning strategies. We also show that these risk bounds are tighter than their traditional counterparts such as VC-dimension and Rademacher complexity-based bounds that explicitly take into account the hypothesis class complexity. Finally, we show how these bounds are able to guide the model selection for the set covering machine algorithm enabling it to learn by bound minimization.
About 60% of the software development cost for online applications is related to developing user interfaces commonly used by the end users to interact with those applications. Frequent small changes to user interfaces (UIs) however easily break about 70% of the test cases intended to mimic users' interactions. Fixing broken tests results in extra costs of maintenance while lessening the benefits of test automation. One of the biggest challenges in creating resilient UI level test cases is to identify and locate the elements of the UI in a way that small UI changes do not break the way in which an element was originally located.In this paper, we present our early efforts in creating a test development framework that makes test cases independent of the internal structure of the UIs, so that a change in the structure does not break any test as long as the functionalities validated in the test are not changed. Our approach is inspired by the way human interact with the UIs of online applications and how those interactions are described and communicated in natural language to others. Visual landmarks such as texts (via a perceptual activity) help end users to locate their points of interest during their interaction with the application in a way that is independent of the internal structure of application.Our framework could significantly reduce the cost of test maintenance by enabling software engineers create UI level tests that are naturally resistant to UI changes.
Une des objectifs poursuivis par la recherche en apprentissage automatique est la construction de bons classificateurs à partir d'un ensemble d'exemples étiquetés. Certains problèmes nécessitent de réunir un grand ensemble d'exemples étiquetés, ce qui peut s'avérer long et coûteux. Afin de réduire ces efforts, il est possible d'utiliser les algorithmes d'apprentissage actif. Ces algorithmes tirent profit de la possibilité de faire quelques demandes d'étiquetage parmi un grand ensemble d'exemples non-étiquetés pour construire un classificateur précis. Il est cependant important de préciser que les algorithmes d'apprentissage actif actuels possèdent eux-mêmes quelques points faibles connus qui peuvent les mener à performer inadéquatement dans certaines situations. Dans cette thèse, nous proposons un nouvel algorithme d'apprentissage actif. Notre algorithme atténue certains points faibles des précédents algorithmes d'apprentissage actif, et il se révèle trés compétitif aux algorithmes d'apprentissage actif bien-connus. De plus, notre algorithme est facile à implémenter.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.