In this paper, we present the potential of Explainable Artificial Intelligence methods for decisi... more In this paper, we present the potential of Explainable Artificial Intelligence methods for decision support in medical image analysis scenarios. Using three types of explainable methods applied to the same medical image data set, we aimed to improve the comprehensibility of the decisions provided by the Convolutional Neural Network (CNN). In vivo gastral images obtained by a video capsule endoscopy (VCE) were the subject of visual explanations, with the goal of increasing health professionals’ trust in black-box predictions. We implemented two post hoc interpretable machine learning methods, called Local Interpretable Model-Agnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP), and an alternative explanation approach, the Contextual Importance and Utility (CIU) method. The produced explanations were assessed by human evaluation. We conducted three user studies based on explanations provided by LIME, SHAP and CIU. Users from different non-medical backgrounds carried o...
Academics have recently emphasised the need for research in operations management to improve in n... more Academics have recently emphasised the need for research in operations management to improve in novelty and its relevance to practice. However, explorative research in innovative practice is often time consuming and risky for academic researchers; as the quality of eventual research results is uncertain. We argue that the difficulty of improving novelty and relevance stems from a lack of criteria for evaluating research-in-process, especially in situations where the design and application of new types of solutions changes operational practices. Developing such evaluation criteria for problem solving research would facilitate the early recognition of novel and relevant research problems, and enable researchers to better discriminate irrelevant or dysfunctional concepts. This study identifies important gaps in the evaluation of research in operations management and draws upon prior contributions of explorative and evaluative explorative research, to establish a framework for analysing research-in-process. Based on the analysis of current research activity, a set of evaluative criteria is proposed for different stages of a research process.
The use of neural networks is still difficult in many application areas due to the lack of explan... more The use of neural networks is still difficult in many application areas due to the lack of explanation facilities (the « black box » problem). The concepts of contextual importance and contextual utility presented make it possible to explain the results of neural networks in a user-understandable way. The explanations obtained are of same quality as those of expert systems, but they may be more flexible since the reasoning module and the explanation module are completely separated. The numerical complexity of estimating the contextual importance and contextual utility is to a great extent solved by the neural net proposed (INKA), which also has good function approximation and training properties.
The eligibility trace is one of the most used mechanisms to speed up reinforcement learning. Earl... more The eligibility trace is one of the most used mechanisms to speed up reinforcement learning. Earlier reported experiments seem to indicate that replacing eligibility traces would perform better than accumulating eligibility traces. However, replacing traces are currently not applicable when using function approximation methods where states are not represented uniquely by binary values. This paper proposes two modifications to replacing traces that overcome this limitation. Experimental results from the Mountain-Car task indicate that the new replacing traces outperform both the accumulating and the 'ordinary' replacing traces.
The eligibility trace is one of the most used mechanisms to speed up reinforcement learning. Earl... more The eligibility trace is one of the most used mechanisms to speed up reinforcement learning. Earlier reported experiments seem to indicate that replacing eligibility traces would perform better than accumulating eligibility traces. However, replacing traces are currently not applicable when using function approximation methods where states are not represented uniquely by binary values. This paper proposes two modifications to replacing traces that overcome this limitation. Experimental results from the Mountain-Car task indicate that the new replacing traces outperform both the accumulating and the 'ordinary' replacing traces.
The eligibility trace is one of the most used mechanisms to speed up reinforcement learning. Earl... more The eligibility trace is one of the most used mechanisms to speed up reinforcement learning. Earlier reported experiments seem to indicate that replacing eligibility traces would perform better than accumulating eligibility traces. However, replacing traces are currently not applicable when using function approximation methods where states are not represented uniquely by binary values. This paper proposes two modifications to replacing traces that overcome this limitation. Experimental results from the Mountain-Car task indicate that the new replacing traces outperform both the accumulating and the 'ordinary' replacing traces.
Multivariate time series with missing data is ubiquitous when the streaming data is collected by ... more Multivariate time series with missing data is ubiquitous when the streaming data is collected by sensors or any other recording instruments. For instance, the outdoor sensors gathering different meteorological variables may encounter low material sensitivity to specific situations, leading to incomplete information gathering. This is problematic in time series prediction with massive missingness and different missing rate of variables. Contribution addressing this problem on the regression task of meteorological datasets by employing Long Short-Term Memory (LSTM), capable of controlling the information flow with its memory unit, is still missing. In this paper, we propose a novel model called forward and backward variable-sensitive LSTM (FBVS-LSTM) consisting of two decay mechanisms and some informative data. The model inputs are mainly the missing indicator, time intervals of missingness in both forward and backward direction and missing rate of each variable. We employ this inform...
The evolution of Internet of Things (IoT) technology has led to an increased emphasis on edge com... more The evolution of Internet of Things (IoT) technology has led to an increased emphasis on edge computing for Cyber-Physical Systems (CPS), in which applications rely on processing data closer to the data sources, and sharing the results across heterogeneous clusters. This has simplified the data exchanges between IoT/CPS systems, the cloud, and the edge for managing low latency, minimal bandwidth, and fault-tolerant applications. Nonetheless, many of these applications administer data collection on the edge and offer data analytic and storage
The Internet of Things (IoT) has promised a future where everything gets connected. Unfortunately... more The Internet of Things (IoT) has promised a future where everything gets connected. Unfortunately, building a single global ecosystem of Things that communicate with each other seamlessly is virtually impossible today. The reason is that the IoT is essentially a collection of isolated "Intranets of Things", also referred to as "vertical silos", which cannot easily and efficiently interact with each other. Smart cities are perhaps the most striking examples of this problem since they comprise a wide range of stakeholders and service providers who must work together, including urban planners, financial organisations, public and private service providers, telecommunication providers, industries, citizens, and so forth. Within this context, the contribution of this paper is threefold: (i) discuss business and technological implications as well as challenges of creating successful open innovation ecosystems, (ii) present the technological building blocks underlying an...
Traditionally, production control on construction sites has been challenging, and still remains c... more Traditionally, production control on construction sites has been challenging, and still remains challenging. The ad-hoc production control methods that are usually used, most of which are informal, foster uncertainty that prevents smooth production flow. Lean construction methods such as the Last Planner System have partially tackled this problem by involving site teams into the decision making process and having them report back to the production management system. However, such systems have relatively long "lookahead" planning cycles to respond to the dynamic production requirements of construction, where daily, if not hourly control is needed. New solutions have been proposed such as VisiLean, KanBIM, etc., but again these types of construction management systems require the proximity and availability of computer devices to workers. Through this paper, the authors investigate how the communication framework underlying such construction management systems can be further improved so as to fully or partially automate various communication functions across the construction project lifecycle (e.g., to enable lean and close to real-time reporting of production control information). To this end, the present paper provides evidences of how the Internet of Things (IoT) and related standards can contribute to such an improvement. The paper then provides first insights-through various construction scenarios-into how the proposed communication framework can be beneficial for various actors and core business perspectives, from lean construction management to the management of the entire building lifecycle.
Proceedings of 4th International Conference on Data Management Technologies and Applications, 2015
Businesses are increasingly using their enterprise data for strategic decision-making activities.... more 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.
The International Journal of Logistics Management, 2015
Purpose – The purpose of this paper is to propose a typology of radio frequency identification (R... more Purpose – The purpose of this paper is to propose a typology of radio frequency identification (RFID)-based tracking solution designs to fit differing fashion supply chains. The typology is presented as principles of form and function contributing toward a design theory of configurable RFID tracking for fashion logistics. Design/methodology/approach – The typology is developed based on a case study of a logistics service provider (LSP) interested in designing a tracking solution for different customers in fashion logistics. In addition to the LSP, four fashion retailers were involved in the study. The case study was carried out using a review of existing RFID tracking implementations in the fashion industry, analysis of an RFID tracking pilot conducted by the case company, and interviews with representatives of the retailers. Findings – By varying three design parameters (place of tagging, place of tracking start and place of tracking end) a tracking solution can be configured to fi...
The context where European manufacturers of industrial systems operate has dramatically changed o... more The context where European manufacturers of industrial systems operate has dramatically changed over recent years: the pressure of emerging countries they have to face, policy makers' environmental laws and industrial companies' interests are pushing towards sustainable manufacturing and a holistic view of industrial systems. Designers and system engineers are the main actors involved, because they have high influence on product life cycle costs and environmental impacts. However they need tools to pursue a holistic view. The aim of this paper is to propose a closed loop framework to improve life cycle performances of industrial systems, focusing on the automotive sector.
Research on Computational Intelligence for Co-Learning Enabled Healthy and Sustainable Housing
ABSTRACT In this paper we propose a unique idea to create a state-of-the-art adaptive building au... more ABSTRACT In this paper we propose a unique idea to create a state-of-the-art adaptive building automation system. The goal can be achieved by combining the control points from building automation obtained data mining and machine learning as well as residents’ feedback. The system to be developed fulfills the needs of a new kind of building automation system improving real-time housing energy efficiency, healthiness, safety and security, as well as comfort considering residents’ lifestyle.
Guiding exploration by pre-existing knowledge without modifying reward
Neural Networks, 2007
Reinforcement learning is based on exploration of the environment and receiving reward that indic... more Reinforcement learning is based on exploration of the environment and receiving reward that indicates which actions taken by the agent are good and which ones are bad. In many applications receiving even the first reward may require long exploration, during which the agent has no information about its progress. This paper presents an approach that makes it possible to use pre-existing knowledge about the task for guiding exploration through the state space. Concepts of short- and long-term memory combine guidance by pre-existing knowledge with reinforcement learning methods for value function estimation in order to make learning faster while allowing the agent to converge towards a good policy.
In this framework, we create an approach to analyse coupling and openness of state-of-the-art log... more In this framework, we create an approach to analyse coupling and openness of state-of-the-art logistics systems. The basic idea is 1) to understand how different environments have succeeded in supporting the adoption of tracking-based value offerings to consumers, contractors, and designers in the application area, 2) to investigate what mechanisms are used in the state of the art examples to make the environment an open-source, distributed, collaborative, self-adaptive, or otherwise easy-to-use application environment for SMEs. The analysis approach is based on the required effort invested in modifying the implemented services in order to make them function in a newly defined manner, and extending the implemented services to new participants. Also, the required effort for switching to another system or system supplier is considered a valid criterion.
Uploads
Papers by Kary Främling