Abstract

Abstract
Machine Learning problems are significantly growing in complexity, either due to an increase in the volume of data, to new forms of data, or due to the change of data over time. This poses new challenges that are both technical and scientific. In this paper we propose a Distributed Learning System that runs on top of a Hadoop cluster, leveraging its native functionalities. It is guided by the principle of data locality. Data are distributed across the cluster, so models are also distributed and trained in parallel. Models are thus seen as Ensembles of base models, and predictions are made by combining the predictions of the base models. Moreover, models are replicated and distributed across the cluster, so that multiple nodes can answer requests. This results in a system that is both resilient and with high availability. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Abstract
E-auditing has been evolving along with the global phenomenon of digitization. Auditors are dealing with new technological challenges and there is the need for more sophisticated tools to support their activities. However, the digital processes used for identifying and validating inconsistencies in the organizations’ financial information are not very efficient. Due to the high number of violations occurrences of tax e-auditing rules, in which many of them turn out to be irrelevant; the auditors’ work is often hindered, which may lead to incomplete data analysis. In this paper, we propose an approach to the e-auditing process based on the SAF-T (PT) files semantics enrichment using a graph-based data structure representation format. Using a graph-based data representation, we can take advantage of another way to perform queries and discovery mechanisms to retrieve information and knowledge, easing the auditing process and consequently enhancing the outcome of the tax e-auditing rules application. © 2021, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Abstract
This paper proposes a scheme to determine the optical dispersion properties of a medium using multiple localized surface plasmon resonances (SPR) in a D-shaped photonic crystal fiber (PCF) whose flat surface is covered by three adjacent gold layers of different thicknesses. Using computational simulations, we show how to customize plasmon resonances at different wavelengths, thus allowing for obtaining the second-order dispersion. The central aspect of this sensing configuration is to balance miniaturization with low coupling between the different localized plasmon modes in adjacent metallic nanostructures. The determination of the optical dispersion over a large spectral range provides information on the concentration of different constituents of a medium, which is of paramount importance when monitoring media with time-varying concentrations, such as fluidic media.

Abstract
In this paper, we advance a multimodal optimization music mashup creation model for loop recombination at scale. The motivation to pursue such a model is to 1) tackle current scalability limitations in state-of-the-art (brute force) models while enforcing the 2) compatibility, i.e., recombination quality, of audio loops, and 3) a pool of diverse solutions that can accommodate personal user preferences or promote different musical styles. To this end, we adopt the Artificial Immune System (AIS) opt-aiNet algorithm to efficiently compute a population of compatible and diverse mashups from loop recombinations. Optimal mashups result from local minima in a feature space that objectively represents harmonic and rhythmic compatibility. We implemented our model as a prototype application named Mixmash-AIS, and conducted an objective evaluation that tackles three dimensions: loop recombination compatibility, mashups diversity, and computational model efficiency. The conducted evaluation compares the proposed system to a standard genetic algorithm (GA) and a brute force (BF) approach. While the GA stands as the most efficient algorithm, its poor results in terms of compatibility reinforce the primacy of the AIS opt-aiNet in efficiently finding optimal compatible loop mashups. Furthermore, the AIS opt-aiNet showed to promote a diverse mashup population, outperforming both GA or BF approaches. © 2021 Owner/Author.

Abstract
This article presents our recent work on the topic of learning from data streams. We focus on emerging topics, including fraud detection, learning from rare cases, and hyper-parameter tuning for streaming data.