Abstract
The vast majority of studies in meta-learning uses only few performance measures when characterizing different machine learning algorithms. The measure Adjusted Ratios of Ratio (ARR) addresses the problem of how to evaluate the quality of a model based on the accuracy and training time. Unfortunately, this measure suffers from a shortcoming that is described in this paper. A new solution is proposed and it is shown that the proposed function satisfies the criterion of monotonicity, unlike ARR.

Abstract
Sentence Reduction has recently received a great attention from the research community of Automatic Text Summarization. Sentence Reduction consists in the elimination of sentence components such as words, part-of-speech tags sequences or chunks without highly deteriorating the information contained in the sentence and its grammatical correctness. In this paper, we present an unsupervised scalable methodology for learning sentence reduction rules. Paraphrases are first discovered within a collection of automatically crawled Web News Stories and then textually aligned in order to extract interchangeable text fragment candidates, in particular reduction cases. As only positive examples exist, Inductive Logic Programming (ILP) provides an interesting learning paradigm for the extraction of sentence reduction rules. As a consequence, reduction cases are transformed into first order logic clauses to supply a massive set of suitable learning instances and an ILP learning environment is defined within the context of the Aleph framework. Experiments evidence good results in terms of irrelevancy elimination, syntactical correctness and reduction rate in a real-world environment as opposed to other methodologies proposed so far.

Abstract
Social Network Analysis (SNA) is an important research area. It originated in sociology but has spread to other areas of research, including anthropology, biology, information science, organizational studies, political science, and computer science. This has stimulated research on how to support SNA with the development of new algorithms. One of the critical areas involves calculation of different centrality measures. The challenge is how to do this fast, as many increasingly larger datasets are available. Our contribution is an incremental version of the Laplacian Centrality measure that can be applied not only to large graphs but also to dynamically changing networks. We have conducted several tests with different types of evolving networks. We show that our incremental version can process a given large network, faster than the corresponding batch version in both incremental and full dynamic network setups. © Springer International Publishing AG 2018.

Abstract

Abstract

Abstract
Machine learning users need methods that can help them identify algorithms or even workflows (combination of algorithms with preprocessing tasks, using or not hyperparameter configurations that are different from the defaults), that achieve the potentially best performance. Our study was oriented towards average ranking (AR), an algorithm selection method that exploits meta-data obtained on prior datasets. We focused on extending the use of a variant of AR* that takes A3R as the relevant metric (combining accuracy and run time). The extension is made at the level of diversity of the portfolio of workflows that is made available to AR. Our aim was to establish whether feature selection and different hyperparameter configurations improve the process of identifying a good solution. To evaluate our proposal we have carried out extensive experiments in a leave-one-out mode. The results show that AR* was able to select workflows that are likely to lead to good results, especially when the portfolio is diverse. We additionally performed a comparison of AR* with Auto-WEKA, running with different time budgets. Our proposed method shows some advantage over Auto-WEKA, particularly when the time budgets are small.