Umeå universitets logga

umu.sePublikationer
Ändra sökning
Avgränsa sökresultatet
1 - 29 av 29
RefereraExporteraLänk till träfflistan
Permanent länk
Referera
Referensformat
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Annat format
Fler format
Språk
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Annat språk
Fler språk
Utmatningsformat
  • html
  • text
  • asciidoc
  • rtf
Träffar per sida
  • 5
  • 10
  • 20
  • 50
  • 100
  • 250
Sortering
  • Standard (Relevans)
  • Författare A-Ö
  • Författare Ö-A
  • Titel A-Ö
  • Titel Ö-A
  • Publikationstyp A-Ö
  • Publikationstyp Ö-A
  • Äldst först
  • Nyast först
  • Skapad (Äldst först)
  • Skapad (Nyast först)
  • Senast uppdaterad (Äldst först)
  • Senast uppdaterad (Nyast först)
  • Disputationsdatum (tidigaste först)
  • Disputationsdatum (senaste först)
  • Standard (Relevans)
  • Författare A-Ö
  • Författare Ö-A
  • Titel A-Ö
  • Titel Ö-A
  • Publikationstyp A-Ö
  • Publikationstyp Ö-A
  • Äldst först
  • Nyast först
  • Skapad (Äldst först)
  • Skapad (Nyast först)
  • Senast uppdaterad (Äldst först)
  • Senast uppdaterad (Nyast först)
  • Disputationsdatum (tidigaste först)
  • Disputationsdatum (senaste först)
Markera
Maxantalet träffar du kan exportera från sökgränssnittet är 250. Vid större uttag använd dig av utsökningar.
  • 1.
    Ait-Mlouk, Addi
    et al.
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    Jiang, Lili
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    KBot: a Knowledge graph based chatBot for natural language understanding over linked data2020Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 8, s. 149220-149230Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    With the rapid progress of the semantic web, a huge amount of structured data has become available on the web in the form of knowledge bases (KBs). Making these data accessible and useful for end-users is one of the main objectives of chatbots over linked data. Building a chatbot over linked data raises different challenges, including user queries understanding, multiple knowledge base support, and multilingual aspect. To address these challenges, we first design and develop an architecture to provide an interactive user interface. Secondly, we propose a machine learning approach based on intent classification and natural language understanding to understand user intents and generate SPARQL queries. We especially process a new social network dataset (i.e., myPersonality) and add it to the existing knowledge bases to extend the chatbot capabilities by understanding analytical queries. The system can be extended with a new domain on-demand, flexible, multiple knowledge base, multilingual, and allows intuitive creation and execution of different tasks for an extensive range of topics. Furthermore, evaluation and application cases in the chatbot are provided to show how it facilitates interactive semantic data towards different real application scenarios and showcase the proposed approach for a knowledge graph and data-driven chatbot.

    Ladda ner fulltext (pdf)
    fulltext
  • 2.
    Ali, Hazrat
    et al.
    Umeå universitet, Medicinska fakulteten, Institutionen för strålningsvetenskaper, Radiofysik.
    Umander, Johannes
    Umeå universitet, Medicinska fakulteten, Institutionen för strålningsvetenskaper, Radiofysik.
    Rohlén, Robin
    Umeå universitet, Medicinska fakulteten, Institutionen för strålningsvetenskaper, Radiofysik.
    Grönlund, Christer
    Umeå universitet, Medicinska fakulteten, Institutionen för strålningsvetenskaper, Radiofysik.
    A Deep Learning Pipeline for Identification of Motor Units in Musculoskeletal Ultrasound2020Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 8, s. 170595-170608Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    Skeletal muscles are functionally regulated by populations of so-called motor units (MUs). An MU comprises a bundle of muscle fibers controlled by a neuron from the spinal cord. Current methods to diagnose neuromuscular diseases and monitor rehabilitation, and study sports sciences rely on recording and analyzing the bio-electric activity of the MUs. However, these methods provide information from a limited part of a muscle. Ultrasound imaging provides information from a large part of the muscle. It has recently been shown that ultrafast ultrasound imaging can be used to record and analyze the mechanical response of individual MUs using blind source separation. In this work, we present an alternative method - a deep learning pipeline - to identify active MUs in ultrasound image sequences, including segmentation of their territories and signal estimation of their mechanical responses (twitch train). We train and evaluate the model using simulated data mimicking the complex activation pattern of tens of activated MUs with overlapping territories and partially synchronized activation patterns. Using a slow fusion approach (based on 3D CNNs), we transform the spatiotemporal image sequence data to 2D representations and apply a deep neural network architecture for segmentation. Next, we employ a second deep neural network architecture for signal estimation. The results show that the proposed pipeline can effectively identify individual MUs, estimate their territories, and estimate their twitch train signal at low contraction forces. The framework can retain spatio-temporal consistencies and information of the mechanical response of MU activity even when the ultrasound image sequences are transformed into a 2D representation for compatibility with more traditional computer vision and image processing techniques. The proposed pipeline is potentially useful to identify simultaneously active MUs in whole muscles in ultrasound image sequences of voluntary skeletal muscle contractions at low force levels.

    Ladda ner fulltext (pdf)
    fulltext
  • 3.
    Asan, Noor Badariah
    et al.
    Ångström Laboratory, Microwaves in Medical Engineering Group, Department of Electrical Engineering, Uppsala University, Uppsala, Sweden; Centre for Telecommunication Research and Innovation (CeTRI), Fakulti Kejuruteraan Elektronik dan Kejuruteraan Komputer, Universiti Teknikal Malaysia Melaka, Durian Tunggal, Malaysia.
    Hassan, Emadeldeen
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för tillämpad fysik och elektronik. Department of Electronics and Electrical Communications, Menoufia University, Menouf, Egypt; Hannover Centre for Optical Technologies, Cluster of Excellence PhoenixD, Leibniz University Hannover, Hanover, Germany; Faculty of Mechanical Engineering, Institute of Transport and Automation Technology, Leibniz University Hannover, Garbsen, Germany.
    Perez, Mauricio D.
    Ångström Laboratory, Microwaves in Medical Engineering Group, Department of Electrical Engineering, Uppsala University, Uppsala, Sweden.
    Joseph, Laya
    Ångström Laboratory, Microwaves in Medical Engineering Group, Department of Electrical Engineering, Uppsala University, Uppsala, Sweden.
    Berggren, Martin
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    Voigt, Thiemo
    Department of Information Technology, Uppsala University, Uppsala, Sweden.
    Augustine, Robin
    Ångström Laboratory, Microwaves in Medical Engineering Group, Department of Electrical Engineering, Uppsala University, Uppsala, Sweden.
    Fat-IntraBody Communication at 5.8 GHz: Verification of Dynamic Body Movement Effects using Computer Simulation and Experiments2021Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 9, s. 48429-48445Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    This paper presents numerical modeling and experimental validation of the signal path loss at the 5.8 GHz Industrial, Scientific, and Medical (ISM) band, performed in the context of fat-intrabody communication (fat-IBC), a novel intrabody communication platform using the body-omnipresent fat tissue as the key wave-guiding medium. Such work extends our previous works at 2.0 and 2.4 GHz in the characterization of its performance in other useful frequency range. In addition, this paper also includes studies of both static and dynamic human body movements. In order to provide with a more comprehensive characterization of the communication performance at this frequency, this work focuses on investigating the path loss at different configurations of fat tissue thickness, antenna polarizations, and locations in the fat channel. We bring more realism to the experimental validation by using excised tissues from porcine cadaver as both their fat and muscle tissues have electromagnetic characteristics similar to those of human with respect to current state-of-art artificial phantom models. Moreover, for favorable signal excitation and reception in the fat-IBC model, we used topology optimized waveguide probes. These probes provide an almost flat response in the frequency range from 3.2 to 7.1 GHz which is higher than previous probes and improve the evaluation of the performance of the fat-IBC model. We also discuss various aspects of real-world scenarios by examining different models, particularly homogeneous multilayered skin, fat, and muscle tissue. To study the effect of dynamic body movements, we examine the impact of misalignment, both in space and in wave polarization, between implanted nodes. We show in particular that the use of fat-IBC techniques can be extended up in frequency to a broadband channel at 5.8 GHz.

    Ladda ner fulltext (pdf)
    fulltext
  • 4. Asan, Noor Badariah
    et al.
    Hassan, Emadeldeen
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    Perez, Mauricio David
    Shah, Syaiful Redzwan Mohd
    Velander, Jacob
    Blokhuis, Taco J.
    Voigt, Thiemo
    Augustine, Robin
    Assessment of Blood Vessel Effect on Fat-Intrabody Communication Using Numerical and Ex-Vivo Models at 2.45 GHZ2019Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 7, s. 89886-89900Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    The potential offered by the intra-body communication (IBC) over the past few years has resulted in a spike of interest for the topic, specifically for medical applications. Fat-IBC is subsequently a novel alternative technique that utilizes fat tissue as a communication channel. This work aimed to identify such transmission medium and its performance in varying blood-vessel systems at 2.45 GHz, particularly in the context of the IBC and medical applications. It incorporated three-dimensional (3D) electromagnetic simulations and laboratory investigations that implemented models of blood vessels of varying orientations, sizes, and positions. Such investigations were undertaken by using ex-vivo porcine tissues and three blood-vessel system configurations. These configurations represent extreme cases of real-life scenarios that sufficiently elucidated their principal influence on the transmission. The blood-vessel models consisted of ex-vivo muscle tissues and copper rods. The results showed that the blood vessels crossing the channel vertically contributed to 5.1 dB and 17.1 dB signal losses for muscle and copper rods, respectively, which is the worst-case scenario in the context of fat-channel with perturbance. In contrast, blood vessels aligned-longitudinally in the channel have less effect and yielded 4.5 dB and 4.2 dB signal losses for muscle and copper rods, respectively. Meanwhile, the blood vessels crossing the channel horizontally displayed 3.4 dB and 1.9 dB signal losses for muscle and copper rods, respectively, which were the smallest losses among the configurations. The laboratory investigations were in agreement with the simulations. Thus, this work substantiated the fat-IBC signal transmission variability in the context of varying blood vessel configurations.

    Ladda ner fulltext (pdf)
    fulltext
  • 5.
    Asim, Muhammad Nabeel
    et al.
    German Research Center for Artificial Intelligence (DFKI), Kaiserslautern, Germany; Department of Computer Science, Tu Kaiserslautern, Kaiserslautern, Germany.
    Malik, Muhammad Imran
    National Center for Artificial Intelligence (NCAI), National University of Sciences and Technology, Islamabad, Pakistan.
    Zehe, Christoph
    Sartorius Corporate Research, Sartorius Stedim Cellca GmbH, Ulm, Germany.
    Trygg, Johan
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Kemiska institutionen. Sartorius Corporate Research, Sartorius Stedim Data Analytics, Umea, Sweden.
    Dengel, Andreas
    German Research Center for Artificial Intelligence (DFKI), Kaiserslautern, Germany; Department of Computer Science, Tu Kaiserslautern, Kaiserslautern, Germany.
    Ahmed, Sheraz
    German Research Center for Artificial Intelligence (DFKI), Kaiserslautern, Germany.
    A robust and precise convnet for small non-coding rna classification (rpc-snrc)2021Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 9, s. 19379-19390, artikel-id 9257352Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    Small non-coding RNAs (ncRNAs) are attracting increasing attention as they are now considered potentially valuable resources in the development of new drugs intended to cure several human diseases. A prerequisite for the development of drugs targeting ncRNAs or the related pathways is the identification and correct classification of such ncRNAs. State-of-the-art small ncRNA classification methodologies use secondary structural features as input. However, such feature extraction approaches only take global characteristics into account and completely ignore co-relative effects of local structures. Furthermore, secondary structure based approaches incorporate high dimensional feature space which is computationally expensive. The present paper proposes a novel Robust and Precise ConvNet (RPC-snRC) methodology which classifies small ncRNAs into relevant families by utilizing their primary sequence. RPC-snRC methodology learns hierarchical representation of features by utilizing positioning and information on the occurrence of nucleotides. To avoid exploding and vanishing gradient problems, we use an approach similar to DenseNet in which gradient can flow straight from subsequent layers to previous layers. In order to assess the effectiveness of deeper architectures for small ncRNA classification, we also adapted two ResNet architectures having a different number of layers. Experimental results on a benchmark small ncRNA dataset show that the proposed methodology does not only outperform existing small ncRNA classification approaches with a significant performance margin of 10% but it also gives better results than adapted ResNet architectures. To reproduce the results Source code and data set is available at https://github.com/muas16/small-non-coding-RNA-classification.

    Ladda ner fulltext (pdf)
    fulltext
  • 6. Babou, Cheikh Saliou Mbacke
    et al.
    Fall, Doudou
    Kashihara, Shigeru
    Taenaka, Yuzo
    Bhuyan, Monowar H.
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    Niang, Ibrahima
    Kadobayashi, Youki
    Hierarchical Load Balancing and Clustering Technique for Home Edge Computing2020Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 8, s. 127593-127607Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    The edge computing system attracts much more attention and is expected to satisfy ultra-low response time required by emerging IoT applications. Nevertheless, as there were problems on latency such as the emerging traffic requiring very sensitive delay, a new Edge Computing system architecture, namely Home Edge Computing (HEC) supporting these real-time applications has been proposed. HEC is a three-layer architecture made up of HEC servers, which are very close to users, Multi-access Edge Computing (MEC) servers and the central cloud. This paper proposes a solution to solve the problems of latency on HEC servers caused by their limited resources. The increase in the traffic rate creates a long queue on these servers, i.e., a raise in the processing time (delay) for requests. By leveraging, based on clustering and load balancing techniques, we propose a new technique called HEC-Clustering Balance. It allows us to distribute the requests hierarchically on the HEC clusters and another focus of the architecture to avoid congestion on a HEC server to reduce the latency. The results show that HEC-Clustering Balance is more efficient than baseline clustering and load balancing techniques. Thus, compared to the HEC architecture, we reduce the processing time on the HEC servers to 19% and 73% respectively on two experimental scenarios.

    Ladda ner fulltext (pdf)
    fulltext
  • 7.
    Bhutto, Adil B.
    et al.
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    Vu, Xuan-Son
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    Elmroth, Erik
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    Tay, Wee Peng
    School of Electrical & Electronics Engineering, Nanyang Technological University, Nanyang, Singapore.
    Bhuyan, Monowar H.
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    Reinforced Transformer Learning for VSI-DDoS Detection in Edge Clouds2022Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 10, s. 94677-94690Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    Edge-driven software applications often deployed as online services in the cloud-to-edge continuum lack significant protection for services and infrastructures against emerging cyberattacks. Very-Short Intermittent Distributed Denial of Service (VSI-DDoS) attack is one of the biggest factor for diminishing the Quality of Services (QoS) and Quality of Experiences (QoE) for users on edge. Unlike conventional DDoS attacks, these attacks live for a very short time (on the order of a few milliseconds) in the traffic to deceive users with a legitimate service experience. To provide protection, we propose a novel and efficient approach for detecting VSI-DDoS attacks using reinforced transformer learning that mitigates the tail latency and service availability problems in edge clouds. In the presence of attacks, the users’ demand for availing ultra-low latency and high throughput services deployed on the edge, can never be met. Moreover, these attacks send very-short intermittent requests towards the target services that enforce longer delays in users’ responses. The assimilation of transformer with deep reinforcement learning accelerates detection performance under adverse conditions by adapting the dynamic and the most discernible patterns of attacks (e.g., multiplicative temporal dependency, attack dynamism). The extensive experiments with testbed and benchmark datasets demonstrate that the proposed approach is suitable, effective, and efficient for detecting VSI-DDoS attacks in edge clouds. The results outperform state-of-the-art methods with 0.9%-3.2% higher accuracy in both datasets.

    Ladda ner fulltext (pdf)
    fulltext
  • 8. De Alcantara Dias, Bruno Martin
    et al.
    Maria Lagana, Armando Antonio
    Justo, Joao Francisco
    Yoshioka, Leopoldo Rideki
    Dias Santos, Max Mauro
    Gu, Zonghua
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för tillämpad fysik och elektronik.
    Model-based development of an engine control module for a spark ignition engine2018Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 6, s. 53638-53649Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    A Spark ignition (SI) engine is a complex, multi-domain component of the vehicle powertrain system. The engine control module (ECM) for an SI engine must achieve both high performance and good fuel efficiency. In this paper, we present a model-based development methodology for an open architecture ECM, addressing the entire development lifecycle including a control algorithm design, parameter calibration, hardware/software implementation, and verification/validation of the final system, both with bench tests on a dynamometer and in a real vehicle on the road. The ECM is able to achieve similar performance as the original proprietary ECM provided by the original equipment manufacturer. Its flexible and modular design enables easy extensibility with new control algorithms, and development of new engine types.

  • 9. Elias Ortega Paredes, Abraham
    et al.
    Nunes, Lauro Roberto
    Dias Santos, Max Mauro
    de Menezes, Leonardo Rodrigues Araujo Xavier
    Silvia Collazos Linares, Kathya
    Francisco Justo, Joao
    Gu, Zonghua
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för tillämpad fysik och elektronik.
    Simulation Performance Enhancement in Automotive Embedded Control Using the Unscented Transform2020Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 8, s. 222041-222049Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    Automotive embedded systems comprise several domains, such as in software, electrical, electronics, and control. When designing and testing functions at the top level, one generally ignores the uncertainties arising from the electrical and electronic effects, which could lead to an irregular behavior and deteriorate their performance even using the appropriate methodology for designing the embedded control systems. Then, the studies and comparison on the effect of uncertainty in the automotive domain are important to improve the overall performance of those control systems. Here, we explored the uncertainty in control systems using the Monte Carlo (MC) and unscented transform (UT) methods. These methods have been applied to a mobile seat platform (MSP) and a light emitting diode (LED) used for lighting of heavy-duty vehicles. The UT for embedded control systems has shown better performance when compared to the Monte Carlo method, in order to reduce the number of required variables and computational resources in the simulation of failures and test-case generation. Finally, this investigation brings another application for the UT, in order to exemplify its applicability and advantages when compared with the other methods.

  • 10.
    Fouladgar, Nazanin
    et al.
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    Alirezaie, Marjan
    Centre for Applied Autonomous Sensor Systems (AASS), Örebro, Sweden.
    Främling, Kary
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap. Aalto University, School of Science and Technology, Finland.
    Metrics and Evaluations of Time Series Explanations: An Application in Affect Computing2022Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 10, s. 23995-24009Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    Explainable artificial intelligence (XAI) has shed light on enormous applications by clarifying why neural models make specific decisions. However, it remains challenging to measure how sensitive XAI solutions are to the explanations of neural models. Although different evaluation metrics have been proposed to measure sensitivity, the main focus has been on the visual and textual data. There is insufficient attention devoted to the sensitivity metrics tailored for time series data. In this paper, we formulate several metrics, including max short-term sensitivity (MSS) , max long-term sensitivity (MLS) , average short-term sensitivity (ASS) and average long-term sensitivity (ALS) , that target the sensitivity of XAI models with respect to the generated and real time series. Our hypothesis is that for close series with the same labels, we obtain similar explanations. We evaluate three XAI models, LIME, integrated gradient (IG), and SmoothGrad (SG), on CN-Waterfall, a deep convolutional network. This network is a highly accurate time series classifier in affect computing. Our experiments rely on data- , metric- and XAI hyperparameter- related settings on the WESAD and MAHNOB-HCI datasets. The results reveal that (i) IG and LIME provide a lower sensitivity scale than SG in all the metrics and settings, potentially due to the lower scale of important scores generated by IG and LIME, (ii) the XAI models show higher sensitivities for a smaller window of data, (iii) the sensitivities of XAI models fluctuate when the network parameters and data properties change, and (iv) the XAI models provide unstable sensitivities under different settings of hyperparameters.

    Ladda ner fulltext (pdf)
    fulltext
  • 11. Javed, Asad
    et al.
    Kubler, Sylvain
    Malhi, Avleen
    Nurminen, Antti
    Robert, Jeremy
    Främling, Kary
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap. Department of Computer Science, Aalto University, Espoo, Finland.
    bIoTope: Building an IoT Open Innovation Ecosystem for Smart Cities2020Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 8, s. 224318-224342Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    The Internet of Things (IoT) has led towards a digital world in which everything becomes connected. Unfortunately, most of the currently marketed connected devices feed vertically-oriented closed systems (commonly referred to as vertical silos) which prevent the development of a unified global IoT. This issue is all the more valid in complex environments, such as smart cities, in which exceedingly large amounts of heterogeneous sensor data are collected, and in which platforms and stakeholders should also be able to interact and cooperate. Therefore, it is of utmost importance to move towards the creation of open IoT ecosystems to support efficient smart city service integration, discovery and composition. This paper contributes to the specifications of such an ecosystem, which has been developed as part of the EU's H2020 bIoTope project. The novelty of this ecosystem compared with the current literature is threefold: (i) it is based on the extensive use of open communication and data standards, notably O-MI and O-DF standards, that foster technical, syntactic and semantic interoperability over domains; (ii) it proposes an innovative service marketplace for data/service publication, discovery and incentivization; (iii) it integrates security functionalities at the IoT gateway level. The practicability of our ecosystem has been validated through several smart city proofs-of-concept set up in three distinct cities: Helsinki, Lyon and Brussels. Given the five major themes defined in the CITYKeys (a smart city performance indicator framework), namely People, Planet, Prosperity, Governance and Propagation, bIoTope mainly contributes to Prosperity-related metrics, as discussed in this paper.

    Ladda ner fulltext (pdf)
    fulltext
  • 12.
    Javed, Asad
    et al.
    Aalto University.
    Malhi, Avleen
    Aalto University.
    Kinnunen, Tuomas
    Aalto University.
    Främling, Kary
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap. Department of Computer Science, Aalto University, 02150 Espoo, Finland.
    Scalable IoT Platform for Heterogeneous Devices in Smart Environments2020Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 8, s. 211973-211985Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    The Internet of Things (IoT) is envisioned as a ubiquitous computing infrastructure in which everything becomes connected, enabling gigantic information exchange among Things and people. These connected smart Things generate an enormous amount of data which need to be efficiently managed to form a unified global IoT. Unfortunately, due to the lack of acceptable open standards, communication protocols, and support for device/service discovery, the recent IoT deployments in smart environments (e.g., smart home, smart building, smart city) are posing imperative challenges related to interoperability, discovery, and the configuration of deployed objects, since the number of objects is expected to grow over time. Therefore, it is of utmost importance to provide open and scalable solutions for the discovery of devices (i.e., Things), their configuration, and data management. This paper introduces an open and scalable IoT platform by adopting the modular characteristics of edge computing for smart environments. This paper: (i) performs a systematic literature review of IoT-based infrastructures and analyzes the scalability requirements; (ii) proposes a layered IoT platform for smart environments that fosters heterogeneity, interoperability, discovery, and scalability; and (iii) demonstrates the applicability of the proposed solution by relying on a comprehensive study of a Väre smart building use case at Aalto University.

    Ladda ner fulltext (pdf)
    fulltext
  • 13.
    Khairova, Nina
    et al.
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap. National Technical University, Kharkiv Polytechnic Institute, Department of Intelligent Computer Systems, Kharkiv, Ukraine.
    Mamyrbayev, Orken
    Institute of Information and Computational Technologies, Almaty, Kazakhstan.
    Rizun, Nina
    Gdańsk University of Technology, Department of Informatics in Management, Gdańsk, Poland.
    Razno, Mariia
    Friedrich Schiller University Jena, Institut für Slawistik und Kaukasusstudien, Jena, Germany.
    Galiya, Ybytayeva
    Satbayev University, Information Processing and Storage, Department of Cybersecurity, Almaty, Kazakhstan.
    A parallel corpus-based approach to the crime event extraction for low-resource languages2023Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 11, s. 54093-54111Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    These days, a lot of crime-related events take place all over the world. Most of them are reported in news portals and social media. Crime-related event extraction from the published texts can allow monitoring, analysis, and comparison of police or criminal activities in different countries or regions. Existing approaches to event extraction mainly suggest processing texts in English, French, Chinese, and some other resource-rich and well-annotated languages. This paper presents a parallel corpus-based approach that follows a closed-domain event extraction methodology to event extraction from web news articles in low-resource languages. To identify the event, its arguments, and the arguments' roles in the source-language part of the corpus we utilize an enhanced pattern-based method that involves the multilingual synonyms dictionary with knowledge about crime-related concepts and logic-linguistic equations. The event extraction from the target-language part of the corpus uses a cross-lingual crime-related event extraction transfer technique that is based on supplementary knowledge about the semantic similarity patterns of the considered pair of languages. The presented approach does not require a preliminarily annotated corpus for training making it more attractive to low-resource languages and allows extracting TRANSFER, CRIME, and POLICE types of events and their seven subtypes from various topics of news articles simultaneously. Implementation of our approach for the Russian-Kazakh parallel corpus of news portals articles allowed obtaining the F1-measure of crime-related event extraction of over 82% for the source language and 63% for the target language.

    Ladda ner fulltext (pdf)
    fulltext
  • 14. Kleyko, Denis
    et al.
    Osipov, Evgeny
    Wiklund, Urban
    Umeå universitet, Medicinska fakulteten, Institutionen för strålningsvetenskaper, Radiofysik.
    A Hyperdimensional Computing Framework for Analysis of Cardiorespiratory Synchronization During Paced Deep Breathing2019Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 7, s. 34403-34415Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    Autonomic function during deep breathing (DB) is normally scored based on the assumption that the heart rate is synchronized with the breathing. We have observed individuals with subtle arrhythmias during DB, where an autonomic function cannot be evaluated. This paper presents a novel method for analyzing cardiorespiratory synchronization: feature-based analysis of the similarity between heart rate and respiration using the principles of hyperdimensional computing. Heart rate and respiration signals were modeled using Fourier series analysis. Three feature variables were derived and mapped to binary vectors in a high-dimensional space. Using both synthesized data and recordings from patients/healthy subjects, the similarity between the feature vectors was assessed using Hamming distance (high-dimensional space), Euclidean distance (original space), and with a coherence-based index. Methods were evaluated via the classification of the similarity indices into three groups. The distance-based methods achieved good separation of signals into classes with different degrees of cardiorespiratory synchronization, also providing identification of patients with low cardiorespiratory synchronization but high values of conventional DB scores. Moreover, binary high-dimensional vectors allowed an additional analysis of the obtained Hamming distance. Feature-based similarity analysis using hyperdimensional computing is capable of identifying signals with low cardiorespiratory synchronization during DB due to arrhythmias. Vector-based similarity analysis could be applied to other types of feature variables than based on spectral analysis. The proposed methods for robustly assessing cardiorespiratory synchronization during DB facilitate the identification of individuals where the evaluation of the autonomic function is problematic or even impossible, thus increasing the correctness of the conventional DB scores.

  • 15.
    Luan, Siyu
    et al.
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för tillämpad fysik och elektronik.
    Gu, Zonghua
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för tillämpad fysik och elektronik.
    Freidovich, Leonid B.
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för tillämpad fysik och elektronik. Department of Information Technologies and AI, Sirius University of Science and Technology, Sochi, Russian Federation.
    Jiang, Lili
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    Zhao, Qingling
    College of Computer Science and Technology, Nanjing University of Science and Technology, Nanjing, China.
    Out-of-Distribution Detection for Deep Neural Networks with Isolation Forest and Local Outlier Factor2021Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 9, s. 132980-132989Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    Deep Neural Networks (DNNs) are extensively deployed in today's safety-critical autonomous systems thanks to their excellent performance. However, they are known to make mistakes unpredictably, e.g., a DNN may misclassify an object if it is used for perception, or issue unsafe control commands if it is used for planning and control. One common cause for such unpredictable mistakes is Out-of-Distribution (OOD) input samples, i.e., samples that fall outside of the distribution of the training dataset. We present a framework for OOD detection based on outlier detection in one or more hidden layers of a DNN with a runtime monitor based on either Isolation Forest (IF) or Local Outlier Factor (LOF). Performance evaluation indicates that LOF is a promising method in terms of both the Machine Learning metrics of precision, recall, F1 score and accuracy, as well as computational efficiency during testing.

    Ladda ner fulltext (pdf)
    fulltext
  • 16.
    Mishra, Tapas Kumar
    et al.
    Srm University, Department of Computer Science Engineering, Amaravati, Andhra Pradesh, India.
    Sahoo, Kshira Sagar
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap. Srm University, Department of Computer Science Engineering, Amaravati, Andhra Pradesh, India.
    Bilal, Muhammad
    Hankuk University of Foreign Studies, Department of Computer Engineering, Yongin-si, South Korea.
    Shah, Sayed Chhattan
    Hankuk University of Foreign Studies, Department of Information and Communication Engineering, Yongin-si, South Korea.
    Mishra, Manas Kumar
    F. M. Autonomous College, Department of Computer Science, Balasore, Odisha, India.
    Adaptive congestion control mechanism to enhance TCP performance in cooperative IoV2023Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 11, s. 9000-9013Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    One of the main causes of energy consumption in Internet of Vehicles (IoV) networks is an ill-designed network congestion control protocol, which results in numerous packet drops, lower throughput, and increased packet retransmissions. In IoV network, the objective to increase network throughput can be achieved by minimizing packets re- transmission and optimizing bandwidth utilization. It has been observed that the congestion control mechanism (i.e., the congestion window) can plays a vital role in mitigating the aforementioned challenges. Thus, this paper present a cross-layer technique to controlling congestion in an IoV network based on throughput and buffer use. In the proposed approach, the receiver appends two bits in the acknowledgment (ACK) packet that describes the status of the buffer space and link utilization. The sender then uses this information to monitor congestion and limit the transmission of packets from the sender. The proposed model has been experimented extensively and the results demonstrate a significantly higher network performance percentage in terms of buffer utilization, link utilization, throughput, and packet loss.

    Ladda ner fulltext (pdf)
    fulltext
  • 17.
    Mohapatra, Debasis
    et al.
    Berhampur (Government), Parala Maharaja Engineering College, Berhampur, Odisha, India.
    Bhoi, Sourav Kumar
    Berhampur (Government), Parala Maharaja Engineering College, Berhampur, Odisha, India.
    Jena, Kalyan Kumar
    Berhampur (Government), Parala Maharaja Engineering College, Berhampur, Odisha, India.
    Sahoo, Kshira Sagar
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    Nayyar, Anand
    Duy Tan University, Faculty of Information Technology, Graduate School, Da Nang, Viet Nam.
    Shah, Mohd Asif
    Bakhtar University, Kabul, Afghanistan.
    Rank-label anonymization for the privacy-preserving publication of a hypergraph structure2022Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 10, s. 118253-118267Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    Social networks are often published in the form of a simple graph. The simple graph representation of a social graph shows the dyadic relationship among the social entities whereas it is unable to efficiently represent the relationship among more than two entities, such as the relationship found in the social groups. This type of relationship is called super-dyadic relationship, and it can be effectively represented by a hypergraph model. This work proposes an anonymization scheme called rank-label anonymization for the privacy-preserving publication of a hypergraph structure. Here, an attack model called rank-label attack is proposed, and an anonymization solution is provided to counter this attack. The percentage of disclosure risk shows that the rank-label attack is stronger than the existing rank attack. We propose a method based on sequential clustering to achieve rank-label anonymization called sequential rank-label anonymization (SA). Another algorithm called greedy rank-label anonymization (GA) is also proposed. The quality of the anonymization solution reported by SA and GA is compared with the help of normalized anonymization cost (NCost). Results show that the NCost reported by SA is less than that of GA for both Adult and MAG-10 datasets. In Adult dataset, approximately 58% and 62% reduction in the average execution time of GA and SA are obtained than that of a general-purpose computing system due to the use of a high-performance computing system. In MAG-10 dataset, this average reduction percentage is reported to be 56% for GA and 53% for SA. The time complexity of SA is found to be O(n4) whereas it is O(n3) in case of GA.

    Ladda ner fulltext (pdf)
    fulltext
  • 18. Munawar, Faizan
    et al.
    Azmat, Shoaib
    Iqbal, Talha
    Grönlund, Christer
    Umeå universitet, Medicinska fakulteten, Institutionen för strålningsvetenskaper, Radiofysik.
    Ali, Hazrat
    Segmentation of Lungs in Chest X-Ray Image Using Generative Adversarial Networks2020Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 8, s. 153535-153545Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    Chest X-ray (CXR) is a low-cost medical imaging technique. It is a common procedure for the identification of many respiratory diseases compared to MRI, CT, and PET scans. This paper presents the use of generative adversarial networks (GAN) to perform the task of lung segmentation on a given CXR. GANs are popular to generate realistic data by learning the mapping from one domain to another. In our work, the generator of the GAN is trained to generate a segmented mask of a given input CXR. The discriminator distinguishes between a ground truth and the generated mask, and updates the generator through the adversarial loss measure. The objective is to generate masks for the input CXR, which are as realistic as possible compared to the ground truth masks. The model is trained and evaluated using four different discriminators referred to as D1, D2, D3, and D4, respectively. Experimental results on three different CXR datasets reveal that the proposed model is able to achieve a dice-score of 0.9740, and IOU score of 0.943, which are better than other reported state-of-the art results.

    Ladda ner fulltext (pdf)
    fulltext
  • 19.
    Najjar, Amro
    et al.
    ICR/AI Robolab, DCS, University of Luxembourg, Luxembourg.
    Mualla, Yazan
    Connaissance et Intelligence Artificielle Distribuées (CIAD), Université Bourgogne Franche-Comté. Belfort, France.
    Singh, Kamal Deep
    Laboratoire Hubert Curien UMR CNRS 5516, Saint-Étienne, France.
    Picard, Gauthier
    ONERA/DTIS, Université de Toulouse, Toulouse, France.
    Calvaresi, Davide
    Institut Informatique et de Gestion (IIG), University of Applied Sciences and Arts Western Switzerland (HES-SO) Valais-Wallis, Sierre, Switzerland.
    Malhi, Avleen
    Department of Computer Science, Aalto University, Helsinki, Finland.
    Galland, Stéphane
    Connaissance et Intelligence Artificielle Distribuées (CIAD), Université Bourgogne Franche-Comté, UTBM, Belfort, France.
    Främling, Kary
    Department of Computer Science, Aalto University, Helsinki, Finland.
    One-to-Many Negotiation QoE Management Mechanism for End-User Satisfaction2021Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 9, s. 59231-59243Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    Quality of Experience (QoE) is defined as the measure of end-user satisfaction with the service. Most of the existing works addressing QoE-management rely on a binary vision of end-user satisfaction. This vision has been criticized by the growing empirical evidence showing that QoE is rather a degree. This article aims to go beyond the binary vision and propose a QoE management mechanism. We propose a one-to-many negotiation mechanism allowing the provider to undertake satisfaction management: to meet fine-grained user QoE goals, while still minimizing the costs. This problem is formulated as an optimization problem, for which a linear model is proposed. For reference, a generic linear program solver is used to find the optimal solution, and an alternative heuristic algorithm is devised to improve the responsiveness when the system has to scale up with a fast-growing number of users. Both are implemented and experimentally evaluated against state-of-the-art one-to-many negotiation frameworks.

  • 20.
    Rezk, Nesma
    et al.
    Halmstad University.
    Purnaprajna, Madhura
    Amrita School of Engineering: Bangalore, Karnataka, India.
    Nordström, Tomas
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för tillämpad fysik och elektronik.
    Ul-Abdin, Zain
    Halmstad University.
    Recurrent Neural Networks: An Embedded Computing Perspective2020Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 81, nr 1, s. 57967-57996Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    Recurrent Neural Networks (RNNs) are a class of machine learning algorithms used for applications with time-series and sequential data. Recently, there has been a strong interest in executing RNNs on embedded devices. However, difficulties have arisen because RNN requires high computational capability and a large memory space. In this paper, we review existing implementations of RNN models on embedded platforms and discuss the methods adopted to overcome the limitations of embedded systems. We will define the objectives of mapping RNN algorithms on embedded platforms and the challenges facing their realization. Then, we explain the components of RNN models from an implementation perspective. We also discuss the optimizations applied to RNNs to run efficiently on embedded platforms. Finally, we compare the defined objectives with the implementations and highlight some open research questions and aspects currently not addressed for embedded RNNs. Overall, applying algorithmic optimizations to RNN models and decreasing the memory access overhead is vital to obtain high efficiency. To further increase the implementation efficiency, we point up the more promising optimizations that could be applied in future research. Additionally, this article observes that high performance has been targeted by many implementations, while flexibility has, as yet, been attempted less often. Thus, the article provides some guidelines for RNN hardware designers to support flexibility in a better manner.

    Ladda ner fulltext (pdf)
    fulltext
  • 21.
    Rohlén, Robin
    et al.
    Umeå universitet, Medicinska fakulteten, Institutionen för strålningsvetenskaper, Radiofysik.
    Stålberg, Erik
    Department of Clinical Neurophysiology, Department of Neurosciences, University Hospital, Uppsala University, Sweden.
    Stoverud, Karen-Helene
    Umeå universitet, Medicinska fakulteten, Institutionen för strålningsvetenskaper, Radiofysik.
    Yu, Jun
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för matematik och matematisk statistik.
    Grönlund, Christer
    Umeå universitet, Medicinska fakulteten, Institutionen för strålningsvetenskaper, Radiofysik.
    A Method for Identification of Mechanical Response of Motor Units in Skeletal Muscle Voluntary Contractions using Ultrafast Ultrasound Imaging: Simulations and Experimental Tests2020Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 8, s. 50299-50311Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    The central nervous system coordinates movement through forces generated by motor units (MUs) in skeletal muscles. To analyze MUs function is essential in sports, rehabilitation medicine applications, and neuromuscular diagnostics. The MUs and their function are studied using electromyography. Typically, these methods study only a small muscle volume (1 mm3) or only a superficial (< 1 cm) volume of the muscle. Here we introduce a method to identify so-called mechanical units, i.e., the mechanical response of electrically active MUs, in the whole muscle (4x4 cm, cross-sectional) under voluntary contractions by ultrafast ultrasound imaging and spatiotemporal decomposition. We evaluate the performance of the method by simulation of active MUs' mechanical response under weak contractions. We further test the experimental feasibility on eight healthy subjects. We show the existence of mechanical units that contribute to the tissue dynamics in the biceps brachii at low force levels and that these units are similar to MUs described by electromyography with respect to the number of units, territory sizes, and firing rates. This study introduces a new potential neuromuscular functional imaging method, which could be used to study a variety of questions on muscle physiology that previously were difficult or not possible to address.

    Ladda ner fulltext (pdf)
    fulltext
  • 22.
    Romano, Luigi
    et al.
    Department of Mechanics and Maritime Sciences, Chalmers University of Technology, H&#x00F6;rsalsv&#x00E4;gen 7A, Gothenburg, Sweden.
    Johannesson, Par
    RISE Research Institutes of Sweden, Gibraltargatan 35, Gothenburg, Sweden.
    Nordström, Erik
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för fysik.
    Bruzelius, Fredrik
    Department of Mechanics and Maritime Sciences, Chalmers University of Technology, H&#x00F6;rsalsv&#x00E4;gen 7A, Gothenburg, Sweden.
    Andersson, Rickard
    Volvo AB, Sweden.
    Jacobson, Bengt
    Department of Mechanics and Maritime Sciences, Chalmers University of Technology, H&#x00F6;rsalsv&#x00E4;gen 7A, Gothenburg, Sweden.
    A classification method of road transport missions and applications using the operating cycle format2022Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 10, s. 73087-73121Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    When dealing with customers, original equipment manufacturers (OEMs) classify vehicular usage by resorting to simplified, often colloquial, descriptions that allow for a rough understanding of the operating conditions and the user&#x2019;s needs. In this way, the information retrieved from the customers is exploited to guide their choices in terms of vehicle design and configuration, based on the characteristics of the transport application, labeled using intuitive metrics. However, a common problem in this context is the absence of any formal connection to lower levels of representation that might effectively be used to assess vehicular energy performance in simulation, or for design optimization using mathematical algorithms. Indeed, both processes require more accurate modeling of the surroundings, including exhaustive information about the local road, weather, and traffic conditions. Therefore, starting with a detailed statistical description of the environment, this paper proposes a method for mathematical classification of transport missions and applications within the theoretical framework of the operating cycle (OC). The approach discussed in the paper combines a collection of statistical models structured hierarchically, called a stochastic operating cycle (sOC), with a bird&#x2019;s-eye view description of the operating environment. The latter postulates the existence of different classes, which are representative of the usage and whose definition is based on simple metrics and thresholds expressed mathematically in terms of statistical measures. Algebraic expressions, called operating classes in the paper, are derived analytically for all the stochastic models presented. This establishes a connection between the two levels of representation, enabling to simulate longitudinal vehicle dynamics in virtual environments generated based on the characteristics of the intended application, using log data collected from vehicles and/or information provided by customers. Additionally, the relationships between the two descriptions are formalized using elementary probability operators, allowing for an intuitive characterization of a transport operation. An example is adduced to illustrate a possible application of the proposed method, using six sOCs parametrized from log data collected during real-world missions. The proposed approach may facilitate the interaction between OEMs, customers, and road operators, allowing for planning of maintenance, and optimization of transport missions, components, and configurations using standard procedures and routines.

    Ladda ner fulltext (pdf)
    fulltext
  • 23.
    Seo, Eunil
    et al.
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    Elmroth, Erik
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    MadFed: enhancing federated learning with marginal-data model fusion2023Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 11, s. 102669-102680Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    As the demand for intelligent applications at the network edge grows, so does the need for effective federated learning (FL) techniques. However, FL often relies on non-identically and non-independently distributed local datasets across end devices, which could result in considerable performance degradation. Prior solutions, such as model-driven approaches based on knowledge distillation, meta-learning, and transfer learning, have provided some reprieve. However, their performance suffers under heterogeneous local datasets and highly skewed data distributions. To address these challenges, this study introduces the MArginal Data fusion FEDerated Learning (MadFed) approach, a groundbreaking fusion of model- and data-driven methodologies. By utilizing marginal data, MadFed mitigates data distribution skewness, improves the maximum achievable accuracy, and reduces communication costs. Furthermore, the study demonstrates that the fusion of marginal data can significantly improve performance even with minimal data entries, such as a single entry. For instance, it provides up to a 15.4% accuracy increase and 70.4% communication cost savings when combined with established model-driven methodologies. Conversely, relying solely on these model-driven methodologies can result in poor performance, especially with highly skewed datasets. Significantly, MadFed extends its effectiveness across various FL algorithms and offers a unique method to augment label sets of end devices, thereby enhancing the utility and applicability of federated learning in real-world scenarios. The proposed approach is not only efficient but also adaptable and versatile, promising broader application and potential for widespread adoption in the field.

    Ladda ner fulltext (pdf)
    fulltext
  • 24.
    Testi, Matteo
    et al.
    Integrated Research Centre, Università Campus Bio-Medico di Roma, Rome, Italy; DeepLearningItalia, Bergamo, Italy.
    Ballabio, Matteo
    DeepLearningItalia, Bergamo, Italy.
    Frontoni, Emanuele
    VRAI Laboratory, Department of Political Sciences Communication and International Relations, Università Degli Studi di Macerata, Macerata, Italy.
    Iannello, Giulio
    Department of Engineering, Unit of Computer Systems and Bioinformatics, Università Campus Bio-Medico di Roma, Rome, Italy.
    Moccia, Sara
    The BioRobotics Institute, Scuola Superiore Sant-Anna, Pisa, Italy; Department of Excellence in Robotics and AI, Scuola Superiore Sant-Anna, Pisa, Italy.
    Soda, Paolo
    Umeå universitet, Medicinska fakulteten, Institutionen för strålningsvetenskaper, Radiofysik. Department of Engineering, Unit of Computer Systems and Bioinformatics, Università Campus Bio-Medico di Roma, Rome, Italy.
    Vessio, Gennaro
    Department of Computer Science, Università Degli Studi di Bari Aldo Moro, Bari, Italy.
    MLOps: A Taxonomy and a Methodology2022Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 10, s. 63606-63618Artikel, forskningsöversikt (Refereegranskat)
    Abstract [en]

    Over the past few decades, the substantial growth in enterprise-data availability and the advancements in Artificial Intelligence (AI) have allowed companies to solve real-world problems using Machine Learning (ML). ML Operations (MLOps) represents an effective strategy for bringing ML models from academic resources to useful tools for solving problems in the corporate world. The current literature on MLOps is still mostly disconnected and sporadic. In this work, we review the existing scientific literature and we propose a taxonomy for clustering research papers on MLOps. In addition, we present methodologies and operations aimed at defining an ML pipeline to simplify the release of ML applications in the industry. The pipeline is based on ten steps: business problem understanding, data acquisition, ML methodology, ML training & testing, continuous integration, continuous delivery, continuous training, continuous monitoring, explainability, and sustainability. The scientific and business interest and the impact of MLOps have grown significantly over the past years: the definition of a clear and standardized methodology for conducting MLOps projects is the main contribution of this paper.

    Ladda ner fulltext (pdf)
    fulltext
  • 25.
    Tortora, Matteo
    et al.
    Campus Bio-Medico University of Rome, Unit of Computer Systems and Bioinformatics, Department of Engineering, Rome, Italy.
    Cordelli, Ermanno
    Campus Bio-Medico University of Rome, Unit of Computer Systems and Bioinformatics, Department of Engineering, Rome, Italy.
    Sicilia, Rosa
    Campus Bio-Medico University of Rome, Unit of Computer Systems and Bioinformatics, Department of Engineering, Rome, Italy.
    Nibid, Lorenzo
    Campus Bio-Medico University of Rome, Unit of Anatomical Pathology, Department of Medicine, Rome, Italy.
    Ippolito, Edy
    Campus Bio-Medico University of Rome, Unit of Radiation Oncology, Department of Medicine, Rome, Italy.
    Perrone, Giuseppe
    Campus Bio-Medico University of Rome, Unit of Anatomical Pathology, Department of Medicine, Rome, Italy.
    Ramella, Sara
    Campus Bio-Medico University of Rome, Unit of Radiation Oncology, Department of Medicine, Rome, Italy.
    Soda, Paolo
    Umeå universitet, Medicinska fakulteten, Institutionen för strålningsvetenskaper, Radiofysik. Campus Bio-Medico University of Rome, Unit of Computer Systems and Bioinformatics, Department of Engineering, Rome, Italy.
    RadioPathomics: multimodal learning in non-small cell lung cancer for adaptive radiotherapy2023Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 11, s. 47563-47578Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    Current practice in cancer treatment collects multimodal data, such as radiology images, histopathology slides, genomics and clinical data. The importance of these data sources taken individually has fostered the recent rise of radiomics and pathomics, i.e., the extraction of quantitative features from radiology and histopathology images collected to predict clinical outcomes or guide clinical decisions using artificial intelligence algorithms. Nevertheless, how to combine them into a single multimodal framework is still an open issue. In this work, we develop a multimodal late fusion approach that combines hand-crafted features computed from radiomics, pathomics and clinical data to predict radiotherapy treatment outcomes for non-small-cell lung cancer patients. Within this context, we investigate eight different late fusion rules and two patient-wise aggregation rules leveraging the richness of information given by CT images, whole-slide scans and clinical data. The experiments in leave-one-patient-out cross-validation on an in-house cohort of 33 patients show that the proposed fusion-based multimodal paradigm, with an AUC equal to 90.9%, outperforms each unimodal approach, suggesting that data integration can advance precision medicine. The results also show that late fusion favourably compares against early fusion, another commonly used multimodal approach. As a further contribution, we explore the chance to use a deep learning framework against hand-crafted features. In our scenario characterised by different modalities and a limited amount of data, as it may happen in different areas of cancer research, the results show that the latter is still a viable and effective option for extracting relevant information with respect to the former.

    Ladda ner fulltext (pdf)
    fulltext
  • 26.
    Tronchin, Lorenzo
    et al.
    University Campus Bio-Medico of Rome, Unit of Computer Systems and Bioinformatics, Department of Engineering, Rome, Italy.
    Cordelli, Ermanno
    University Campus Bio-Medico of Rome, Unit of Computer Systems and Bioinformatics, Department of Engineering, Rome, Italy.
    Celsi, Lorenzo Ricciardi
    Elis Innovation Hub, Rome, Italy.
    MacCagnola, Daniele
    Assicurazioni Generali Italia, Advanced Analytics, Milan, Italy.
    Natale, Massimo
    Assicurazioni Generali Italia, Advanced Analytics, Milan, Italy.
    Soda, Paolo
    Umeå universitet, Medicinska fakulteten, Institutionen för diagnostik och intervention. University Campus Bio-Medico of Rome, Unit of Computer Systems and Bioinformatics, Department of Engineering, Rome, Italy.
    Sicilia, Rosa
    University Campus Bio-Medico of Rome, Unit of Computer Systems and Bioinformatics, Department of Engineering, Rome, Italy.
    Translating image XAI to multivariate time series2024Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 12, s. 27484-27500Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    As Artificial Intelligence (AI) is becoming part of our daily lives, the need to understand and trust its decisions is becoming a pressing issue. EXplainable AI (XAI) aims at answering this demand, providing tools to get insights into the models' behaviour and reasoning. Following this trend, our research paper explores the explainability of a deployed multimodal architecture applied to a real-world dataset of multivariate time series. The study aims to enhance the trustworthiness of an AI agent responsible for crash detection in an insurance company's automatic assistance service. By introducing an XAI layer, we provide insights into the AI agent's decision-making process, enabling the optimization of emergency medical services allocation. The dataset consists of real-world telematics data collected from vehicles equipped with black box technology. The challenge lies in explaining the complex interactions within the multivariate time series data to accurately understand the forces applied to vehicles during accidents. To this end, we adapt to this context two state-of-the-art XAI model-specific approaches, originally designed for images. We offer a qualitative and a quantitative evaluation, also comparing with a well-known agnostic method, and further validating our findings on an external dataset. The results show that Integrated Gradients, among the methodologies examined, is the most effective approach. Its ability to handle the complexity of the data provides the most comprehensive and insightful explanations for the considered use case. The findings emphasize the potential of XAI to enhance the trustworthiness of AI systems and optimize emergency response in the insurance industry. Code is available at https://github.com/ltronchin/translating-xai-mts.git.

    Ladda ner fulltext (pdf)
    fulltext
  • 27. Vreča, Jure
    et al.
    Sturm, Karl J. X.
    Gungl, Ernest
    Merchant, Farhad
    Bientinesi, Paolo
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    Leupers, Rainer
    Brezočnik, Zmago
    Accelerating Deep Learning Inference in Constrained Embedded Devices Using Hardware Loops and a Dot Product Unit2020Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 8, s. 165913-165926Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    Deep learning algorithms have seen success in a wide variety of applications, such as machine translation, image and speech recognition, and self-driving cars. However, these algorithms have only recently gained a foothold in the embedded systems domain. Most embedded systems are based on cheap microcontrollers with limited memory capacity, and, thus, are typically seen as not capable of running deep learning algorithms. Nevertheless, we consider that advancements in compression of neural networks and neural network architecture, coupled with an optimized instruction set architecture, could make microcontroller-grade processors suitable for specific low-intensity deep learning applications. We propose a simple instruction set extension with two main components-hardware loops and dot product instructions. To evaluate the effectiveness of the extension, we developed optimized assembly functions for the fully connected and convolutional neural network layers. When using the extensions and the optimized assembly functions, we achieve an average clock cycle count decrease of 73% for a small scale convolutional neural network. On a per layer base, our optimizations decrease the clock cycle count for fully connected layers and convolutional layers by 72% and 78%, respectively. The average energy consumption per inference decreases by 73%. We have shown that adding just hardware loops and dot product instructions has a significant positive effect on processor efficiency in computing neural network functions.

  • 28.
    Wang, Dong
    et al.
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Kemiska institutionen.
    Enlund, Therese
    Mestro AB, Stockholm, Sweden.
    Trygg, Johan
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Kemiska institutionen.
    Tysklind, Mats
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Kemiska institutionen.
    Jiang, Lili
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för datavetenskap.
    Toward Delicate Anomaly Detection of Energy Consumption for Buildings: Enhance the Performance From Two Levels2022Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 10, s. 31649-31659Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    Buildings are highly energy-consuming and therefore are largely accountable for environmental degradation. Detecting anomalous energy consumption is one of the effective ways to reduce energy consumption. Besides, it can contribute to the safety and robustness of building systems since anomalies in the energy data are usually the reflection of malfunctions in building systems. As the most flexible and applicable type of anomaly detection approach, unsupervised anomaly detection has been implemented in several studies for building energy data. However, no studies have investigated the joint influence of data structures and algorithms’ mechanisms on the performance of unsupervised anomaly detection for building energy data. Thus, we put forward a novel workflow based on two levels, data structure level and algorithm mechanism level, to effectively detect the imperceptible anomalies in the energy consumption profiles of buildings. The proposed workflow was implemented in a case study for identifying the anomalies in three real-world energy consumption datasets from two types of commercial buildings. Two aims were achieved through the case study. First, it precisely detected the contextual anomalies concealed beneath the time variation of the energy consumption profiles of the three buildings. The performance in terms of areas under the precision-recall curves (AUC_PR) for the three given datasets were 0.989, 0.941, and 0.957, respectively. Second, more broadly, the joint effect of the two levels was examined. On the data level, all four detectors on the contextualized data were superior to their counterparts on the original data. On the algorithm level, there was a consistent ranking of detectors regarding their detecting performances on the contextualized data. The consistent ranking suggests that local approaches outperform global approaches in the scenarios where the goal is to detect the instances deviating from their contextual neighbors rather than the rest of the entire data.

    Ladda ner fulltext (pdf)
    fulltext
  • 29. Zhang, Ming
    et al.
    Gu, Zonghua
    Umeå universitet, Teknisk-naturvetenskapliga fakulteten, Institutionen för tillämpad fysik och elektronik.
    Zheng, Nenggan
    Ma, De
    Pan, Gang
    Efficient Spiking Neural Networks With Logarithmic Temporal Coding2020Ingår i: IEEE Access, E-ISSN 2169-3536, Vol. 8, s. 98156-98167Artikel i tidskrift (Refereegranskat)
    Abstract [en]

    A Spiking Neural Network (SNN) can be trained indirectly by first training an Artificial Neural Network (ANN) with the conventional backpropagation algorithm, then converting it into an equivalent SNN. To reduce the computational cost of the resulting SNN as measured by the number of spikes, we present Logarithmic Temporal Coding (LTC), where the number of spikes used to encode an activation grows logarithmically with the activation value; and the accompanying Exponentiate-and-Fire (EF) neuron model, which only involves efficient bit-shift and addition operations. Moreover, we improve the training process of ANN to compensate for approximation errors due to LTC. Experimental results indicate that the resulting SNN achieves competitive performance in terms of classification accuracy at significantly lower computational cost than related work.

    Ladda ner fulltext (pdf)
    fulltext
1 - 29 av 29
RefereraExporteraLänk till träfflistan
Permanent länk
Referera
Referensformat
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Annat format
Fler format
Språk
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Annat språk
Fler språk
Utmatningsformat
  • html
  • text
  • asciidoc
  • rtf