Search

Publisher Copyright: © 2021, Springer Nature Switzerland AG. Different explainable AI (XAI) methods are based on different notions of ‘ground truth’. In order to trust explanations of AI systems, the ground truth has to provide fidelity towards the actual behaviour of the AI system. An explanation that has poor fidelity towards the AI system’s actual behaviour can not be trusted no matter how convincing the explanations appear to be for the users. The Contextual Importance and Utility (CIU) method differs from currently popular outcome explanation methods such as Local Interpretable Model-agnostic Explanations (LIME) and Shapley values in several ways. Notably, CIU does not build any intermediate interpretable model like LIME, and it does not make any assumption regarding linearity or additivity of the feature importance. CIU also introduces the value utility notion and a definition of feature importance that is different from LIME and Shapley values. We argue that LIME and Shapley values actually estimate ‘influence’ (rather than ‘importance’), which combines importance and utility. The paper compares the three methods in terms of validity of their ground truth assumption and fidelity towards the underlying model through a series of benchmark tasks. The results confirm that LIME results tend not to be coherent nor stable. CIU and Shapley values give rather similar results when limiting explanations to ‘influence’. However, by separating ‘importance’ and ‘utility’ elements, CIU can provide more expressive and flexible explanations than LIME and Shapley values.

Explaining the result of machine learning models is an active research topic in Artificial Intelligence (AI) domain with an objective to provide mechanisms to understand and interpret the results of the underlying black-box model in a human-understandable form. With this objective, several eXplainable Artificial Intelligence (XAI) methods have been designed and developed based on varied fundamental principles. Some methods such as Local interpretable model agnostic explanations (LIME), SHAP (SHapley Additive exPlanations) are based on the surrogate model while others such as Contextual Importance and Utility (CIU) do not create or rely on the surrogate model to generate its explanation. Despite the difference in underlying principles, these methods use different sampling techniques such as uniform sampling, weighted sampling for generating explanations. CIU, which emphasizes a context-aware decision explanation, employs a uniform sampling method for the generation of representative samples. In this research, we target uniform sampling methods which generate representative samples that do not guarantee to be representative in the presence of strong non-linearities or exceptional input feature value combinations. The objective of this research is to develop a sampling method that addresses these concerns. To address this need, a new adaptive weighted sampling method has been proposed. In order to verify its efficacy in generating explanations, the proposed method has been integrated with CIU, and tested by deploying the special test case

This paper proposes an explainable machine learning tool that can potentially be used for decision support in medical image analysis scenarios. For a decision-support system it is important to be able to reverse-engineer the impact of features on the final decision outcome. In the medical domain, such functionality is typically required to allow applying machine learning to clinical decision making. In this paper, we present initial experiments that have been performed on in-vivo gastral images obtained from capsule endoscopy. Quantitative analysis has been performed to evaluate the utility of the proposed method. Convolutional neural networks have been used for training the validating of the image data set to provide the bleeding classifications. The visual explanations have been provided in the images to help health professionals trust the black box predictions. While the paper focuses on the in-vivo gastral image use case, most findings are generalizable.

IoT systems may provide information from different sensors that may reveal potentially confidential data, such as a person’s presence or not. The primary question to address is how we can identify the sensors and other devices in a reliable way before receiving data from them and using or sharing it. In other words, we need to verify the identity of sensors and devices. A malicious device could claim that it is the legitimate sensor and trigger security problems. For instance, it might send false data about the environment, harmfully affecting the outputs and behavior of the system. For this purpose, using only primary identity values such as IP address, MAC address, and even the public-key cryptography key pair is not enough since IPs can be dynamic, MACs can be spoofed, and cryptography key pairs can be stolen. Therefore, the server requires supplementary security considerations such as contextual features to verify the device identity. This paper presents a measurement-based method to detect and alert false data reports during the reception process by means of sensor behavior. As a proof of concept, we develop a classification-based methodology for device identification, which can be implemented in a real IoT scenario.

Part 5: Industry 4.0 - Digital Twin; International audience; Maintenance is a complicated task that encompasses various activities including fault detection, fault diagnosis, and fault reparation. The advancement of Computer Aided Engineering (CAE) has increased challenges in maintenance as modern assets have became complex mixes of systems and sub systems with complex interaction. Among maintenance activities, fault diagnosis is particularly cumbersome as the reason of failures on the system is often neither obvious in terms of their source nor unique. Early detection and diagnosis of such faults is turning to one of the key requirements for economical and functional asset efficiency. Several methods have been investigated to detect machine faults for a number of years that are relevant for many application domains. In this paper, we present the process history-based method adopting nominal efficiency of Air Handling Unit (AHU) to detect heat recovery failure using Principle Component Analysis (PCA) in combination of the logistic regression method.

openaire: EC/H2020/688203/EU//BIoTope The number of Internet of Things (IoT) vendors is rapidly growing, providing solutions for all levels of the IoT stack. Despite the universal agreement on the need for a standardized technology stack, following the model of the world-wide-web, a large number of industry-driven domain specific standards hinder the development of a single IoT ecosystem. An attempt to solve this challenge is the introduction of O-MI (Open Messaging Interface) and O-DF (Open Data Format), two domain independent standards published by Open Group. Despite their good compatibility, they define no specific security model. This paper takes the first step of defining a security model for these standards by proposing suitable access control and authentication mechanisms that can regulate the rights of different principles and operations defined in these standards. First, a brief introduction is provided of the O-MI and O-DF standards, including a comparison with existing standards. Second, the envisioned security model is presented, together with the implementation details of the plug-in module developed for the O-MI and O-DF reference implementation.

The Industrial Internet promises to radically change and improve many industry's daily business activities, from simple data collection and processing to context-driven, intelligent and pro-active support of workers' everyday tasks and life. The present paper first provides insight into a typical industrial internet application architecture, then it highlights one fundamental arising contradiction: “Who owns the data is often not capable of analyzing it”. This statement is explained by imaging a visionary data supply chain that would realize some of the Industrial Internet promises. To concretely implement such a system, recent standards published by The Open Group are presented, where we highlight the characteristics that make them suitable for Industrial Internet applications. Finally, we discuss comparable solutions and concludes with new business use cases.

Businesses are increasingly using their enterprise data for strategic decision-making activities. In fact, information, derived from data, has become one of the most important tools for businesses to gain competitive edge. Data quality assessment has become a hot topic in numerous sectors and considerable research has been carried out in this respect, although most of the existing frameworks often need to be adapted with respect to the use case needs and features. Within this context, this paper develops a methodology for assessing the quality of enterprises' daily maintenance reporting, relying both on an existing data quality framework and on a Multi-Criteria Decision Making (MCDM) technique. Our methodology is applied in cooperation with a Finnish multinational company in order to evaluate and rank different company sites/office branches (carrying out maintenance activities) according to the quality of their data reporting. Based on this evaluation, the industrial partner wants to establish new action plans for enhanced reporting practices.

Evidence that data quality is hardly considered in expert & intelligent systems is provided.A decision-support tool for maintenance data reporting assessment is developed.A company uses the proposed tool to rank branch offices in terms of reporting quality.Enhanced reporting quality results in enhanced business activities & decision-making. Today's largest and fastest growing companies' assets are no longer physical, but rather digital (software, algorithms...). This is all the more true in the manufacturing, and particularly in the maintenance sector where quality of enterprise maintenance services are closely linked to the quality of maintenance data reporting procedures. If quality of the reported data is too low, it can results in wrong decision-making and loss of money. Furthermore, various maintenance experts are involved and directly concerned about the quality of enterprises' daily maintenance data reporting (e.g., maintenance planners, plant managers...), each one having specific needs and responsibilities. To address this Multi-Criteria Decision Making (MCDM) problem, and since data quality is hardly considered in existing expert maintenance systems, this paper develops a maintenance reporting quality assessment (MRQA) dashboard that enables any company stakeholder to easily - and in real-time - assess/rank company branch offices in terms of maintenance reporting quality. From a theoretical standpoint, AHP is used to integrate various data quality dimensions as well as expert preferences. A use case describes how the proposed MRQA dashboard is being used by a Finnish multinational equipment manufacturer to assess and enhance reporting practices in a specific or a group of branch offices.