In the past recent years, Machine Learning methodologies have been applied in countless application areas. particular, they play a key role enabling Industry 4.0. However, one of main obstacles to diffusion Learning-based applications is related lack interpretability most these methods. this work, we propose an approach for defining `feature importance' Anomaly Detection problems. important task that has enormous applicability industrial scenarios. Indeed, it extremely relevant purpose...

In refrigerators production, vacuum creation is fundamental to guarantee the correct manufacturing of product. Before inserting refrigerant in refrigerator cabinet, tested through a Pirani gauge that assesses pressure within cabinet. Such readings are used evaluate process and verify if leakings present. this work, we employ Deep Learning-based Anomaly Detection approach associate an Score each profile; score can be exploited optimize actions performed by human operators like more detailed...

Gesture Recognition has a prominent importance in smart environment and home automation. Thanks to the availability of Machine Learning approaches it is possible for users define gestures that can be associated with commands environment. In this paper we propose Random Forest-based approach hand movements starting from wireless wearable motion capture data. presented approach, evaluate different feature extraction procedures handle data duration. To enhance reproducibility our results foster...

Anomaly Detection is an unsupervised learning task aimed at detecting anomalous behaviours with respect to historical data. In particular, multivariate has important role in many applications thanks the capability of summarizing status a complex system or observed phenomenon single indicator (typically called `Anomaly Score') and nature that does not require human tagging. The Isolation Forest one most commonly adopted algorithms field Detection, due its proven effectiveness low...

We conduct an audit of pricing algorithms employed by companies in the Italian car insurance industry, primarily gathering quotes through a popular comparison website. While acknowledging complexity we find evidence several problematic practices. show that birthplace and gender have direct sizeable impact on prices quoted to drivers, despite national international regulations against their use. Birthplace, particular, is used quite frequently disadvantage foreign-born drivers born certain...

Deep Learning approaches have revolutionized in the past decade field of Computer Vision and, as a consequence, they are having major impact Industry 4.0 applications like automatic defect classification. Nevertheless, additional data, beside image/video itself, is typically never exploited classification module: this aspect, given abundance data data-intensive manufacturing environments (like semiconductor manufacturing) represents missed opportunity. In work we present use case related to...

One of the main challenges in developing Machine Learning-based solutions for Semiconductor Manufacturing is high number machines production and their differences, even when considering chambers same machine; this poses a challenge scalability context, since development chamber-specific models all equipment fab unsustainable. In work, we present domain adaptation approach Virtual Metrology (VM), one most successful technology context. The provides common VM model two identical-in-design...

Machine Learning-based Anomaly Detection approaches are efficient tools to monitor complex processes. One of the advantages such is that they provide a unique anomaly indicator, quantitative index captures degree 'outlierness' process at hand considering possibly hundreds or more variables same time, typical scenario in semiconductor manufacturing. drawbacks Root Cause Analysis not guided by system itself. In this work, we show effectiveness method, called DIFFI, equip Isolation Forest, one...

We conduct an audit of pricing algorithms employed by companies in the Italian car insurance industry, primarily gathering quotes through a popular comparison website. While acknowledging complexity we find evidence several problematic practices. show that birthplace and gender have direct sizeable impact on prices quoted to drivers, despite national international regulations against their use. Birthplace, particular, is used quite frequently disadvantage foreign-born drivers born certain...

Adversarial Training has proved to be an effective training paradigm enforce robustness against adversarial examples in modern neural network architectures. Despite many efforts, explanations of the foundational principles underpinning effectiveness are limited and far from being widely accepted by Deep Learning community. In this paper, we describe surprising properties adversarially-trained models, shedding light on mechanisms through which attacks is implemented. Moreover, highlight...

Adversarial robustness is one of the most challenging problems in Deep Learning and Computer Vision research. All state-of-the-art techniques require a time-consuming procedure that creates cleverly perturbed images. Due to its cost, many solutions have been proposed avoid Training. However, all these attempts proved ineffective as attacker manages exploit spurious correlations among pixels trigger brittle features implicitly learned by model. This paper first introduces new image filtering...

In the context of human-in-the-loop Machine Learning applications, like Decision Support Systems, interpretability approaches should provide actionable insights without making users wait. this paper, we propose Accelerated Model-agnostic Explanations (AcME), an approach that quickly provides feature importance scores both at global and local level. AcME can be applied a posteriori to each regression or classification model. Not only does compute ranking, but it also what-if analysis tool...