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.

openaire: EC/H2020/688203/EU//BIoTope Industrial Internet of things is becoming a boon to Original Equipment Manufacturers (OEMs) offering after-sales services such as condition-based maintenance and extended warranty for their products. These companies leverage novel digital information infrastructures to improve daily industrial activities, including data collection, remote monitoring and advanced condition-based maintenance services. The emergence of digital infrastructure and new business prospects via servi-tization and quality services encourage companies to collect vast amounts of data that have been generated in different stages of product lifecycles. Despite of the potential benefits, companies are unable to fully harness the opportunities presented by digital information infrastructure because there exist several platforms with variations in technologies and standards resulting in interoperability challenges. This becomes particularly critical when a company sells its products to several clients with different technologies. To overcome such challenges, we investigate the Open Messaging Interface (O-MI) and Open Data Format (O-DF), flexible messaging and data exchange standards that enable seamless integration of different systems. These standards enable interoperability and support time-centric, event-centric, and rate-centric modes of data exchange.

We are entering a new age of AI applications where machine learning is the core technology but machine learning models are generally non-intuitive, opaque and usually complicated for people to unde ...

openaire: EC/H2020/688203/EU//BIoTope Information Systems (ISs) are fundamental to streamline operations and support processes of any modern enterprise. Being able to perform analytics over the data managed in various enterprise ISs is becoming increasingly important for organisational growth. Extract, Transform, and Load (ETL) are the necessary pre-processing steps of any data mining activity. Due to the complexity of modern IS, extracting data is becoming increasingly complicated and time-consuming. In order to ease the process, this paper proposes a methodology and a pilot implementation, that aims to simplify data extraction process by leveraging the end-users’ knowledge and understanding of the specific IS. This paper first provides a brief introduction and the current state of the art regarding existing ETL process and techniques. Then, it explains in details the proposed methodology. Finally, test results of typical data-extraction tasks from 4 commercial ISs are reported.

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.

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.

Service business renewal to support global fleet operations is one of the critical competitiveness factors of technology industries. This paper describes preliminary results of defining an IoT-based framework for fleet management systems (FMS). The main contribution is to provide a systematic outline to the ingredients of the framework and discussing their impact in an industrial setting. The overall aim of the research is to define a technical framework for fleet management solutions and to define the necessary set of functionalities that the framework should support as a base for FMS systems. The framework will be implemented and validated in collaboration with industrial partners.

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.

Building Information Management systems have started to leverage new kinds of digital information infrastructures that integrate activities related to design, budgeting, scheduling, analysis, material management, and human resources. Companies implementing such systems have no other option today than exchanging information among themselves in a proper and efficient way. Industry foundation classes (IFCs) provide useful information structures for such a data sharing. However they do not specify how to capture, exchange and synchronize the information between distinct entities (i.e. information systems like sensors, servers, smart products…) throughout the building lifecycle. Accordingly, this paper investigates existing messaging protocols in order to identify which one is the most appropriate for supporting building lifecycle management; i.e a flexible protocol that provides sufficiently generic communication interfaces. A platform set up on a university campus based on the selected messaging protocol is then presented. Within this context, the research agenda that should be pursued to develop new techniques and algorithms for optimizing the building energy management is announced.

Nowadays, organizations make a point of protecting the confidentiality of their data and assets using firewalls, proxies and NATs, which goes against providing a high level of data usability and interoperability between distinct information systems, or “Things” in the so-called Internet of Things. Such security procedures often prevent two-way communications between nodes located on each side of the firewall. Quantum Lifecycle Management (QLM) messaging has been introduced as a messaging standard proposal that would fulfill the requirements for exchanging the kind of information required by an IoT. In this regard, the QLM piggy backing property proposed in that standard makes it possible to achieve two-way communication through a firewall. This property is introduced in this paper, along with the first proofs-of-concept.

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.

openaire: EC/H2020/688203/EU//BIoTope Many domains are trying to integrate with the Internet of Things (IoT) ecosystem, such as public administrations starting smart city initiatives all over the world. Cities are becoming smart in many ways: smart mobility, smart buildings, smart environment and so on. However, the problem of non-interoperability in the IoT hinders the seamless communication between all kinds of IoT devices. Different domain specific IoT applications use different interoperability standards. These standards are usually not interoperable with each other. IoT applications and ecosystems therefore tend to use a vertical communication model that does not allow data sharing horizontally across different IoT ecosystems. In 2014, The Open Group published two domain-independent IoT messaging standards, O-MI and O-DF, aiming to solve the interoperability problem. In this article we describe the practical use of O-MI/O-DF standards for reaching interoperability in a mobile application for the smart city context, in particular for the Smart Mobility domain, electric vehicle (EV) charging case study. The proof-of-concept of the smart EV charging ecosystem with mobile application user interface was developed as a part of an EU (Horizon 2020) Project bIoTope.