How can we select the best performing data‐driven model? How can we rigorously estimate its generalization error? Statistical learning theory (SLT) answers these questions by deriving nonasymptotic bounds on the generali

How can we select the best performing data‐driven model? How can we rigorously estimate its generalization error? Statistical learning theory (SLT) answers these questions by deriving nonasymptotic bounds on the generalization error of a model or, in other words, by delivering upper bounding of the true error of the learned model based just on quantities computed on the available data. However, for a long time, SLT has been considered only as an abstract theoretical framework, useful for inspiring new learning approaches, but with limited applicability to practical problems. The purpose of this review is to give an intelligible overview of the problems of model selection (MS) and error estimation (EE), by focusing on the ideas behind the different SLT‐based approaches and simplifying most of the technical aspects with the purpose of making them more accessible and usable in practice. We start by presenting the seminal works of the 80s until the most recent results, then discuss open problems and finally outline future directions of this field of research.

This article is categorized under: Technologies > Statistical Fundamentals Algorithmic Development > Statistics

Performance of any search engine relies heavily on its Web crawler. Web crawlers are the programs that get webpages from the Web by following hyperlinks. These webpages are indexed by a search engine and can be

Performance of any search engine relies heavily on its Web crawler. Web crawlers are the programs that get webpages from the Web by following hyperlinks. These webpages are indexed by a search engine and can be retrieved by a user query. In the area of Web crawling, we still lack an exhaustive study that covers all crawling techniques. This study follows the guidelines of systematic literature review and applies it to the field of Web crawling. We used the standard procedure of carrying out a systematic literature review on 248 studies from a total of 1488 articles published in 12 leading journals and other premier conferences and workshops. Existing literature about the Web crawler is classified into different key subareas. Each subarea is further divided according to the techniques being used. We analyzed the distribution of various articles using multiple criteria and depicted conclusions. Various studies that use open source Web crawlers are also reported. We have highlighted future areas of research. We call for an increased awareness in various fields of the Web crawler and identify how techniques from other domains can be used for crawling the Web. Limitations and recommendations for future are also discussed. WIREs Data Mining Knowl Discov 2017, 7:e1218. doi: 10.1002/widm.1218

This article is categorized under: Algorithmic Development > Web Mining Fundamental Concepts of Data and Knowledge > Information Repositories Fundamental Concepts of Data and Knowledge > Motivation and Emergence of Data Mining