Paulo Ferreira

• Rosà, Andrea & Ferreira, Paulo (2026). ICOOOLPS 2025 – Implementation, Compilation, Optimization of Object-Oriented Languages, Programs and Systems Preface. Journal of Object Technology. 25(1), s. 1–2. doi: 10.5381/jot.2026.25.1.e4. Fulltekst i vitenarkiv
• Reinholdt, Jonas Paludan; Jul, Eric Bartley & Ferreira, Paulo (2024). flyDetect: An Android Application for Flight Detection. Sensors. 24(18). doi: 10.3390/s24186158. Fulltekst i vitenarkiv Vis sammendrag

  Over the past years, transport mode recognition has become a large field of research. However, flight as a type of transportation has been mostly overlooked. A system for flight detection might be useful for context-aware applications, but more importantly, it can be used to automatically manage airplane mode on smartphones. Smartphones transmit radio frequency signals which could potentially interfere with aircraft systems, and it is therefore important that devices enable airplane mode to avoid this problem. This paper proposes flyDetect, a method for automatic flight mode detection and an embodiment in the form of an app that demonstrates the viability of the method. Thus, the system uses the accelerometer and barometer in an Android smartphone, can detect the start and end of a flight, and notify other apps or systems on the device when this happens. Our evaluation shows that flyDetect meets the requirements set for the solution, and the results are very promising.

• Kamalian, Mahdieh; Taherkordi, Amirhosein; Payberah, Amir H. & Ferreira, Paulo (2024). FOGFLEET: Fog-Level Federated Transfer Learning for Adaptive Transport Mode Detection. Proceedings of the IEEE. ISSN 0018-9219. s. 22–33. doi: 10.1109/IC2E61754.2024.00010. Fulltekst i vitenarkiv Vis sammendrag

  Transport Mode Detection (TMD) systems play a pivotal role in facilitating applications in transport, urban planning, and more. Exploiting the advancements in smartphone sensing capabilities, TMD systems have evolved for mobile applications, with local classification on smartphones as a common approach. Yet, local approaches relying on centralized training raise privacy concerns due to the transmission of sensitive data (e.g., GPS logs) over the Internet. In this paper, we propose FOGFLEET, a novel Federated Transfer Learning (FTL) framework for TMD, addressing both privacy and performance concerns. Our approach relies on Federated Learning (FL) to train a global model on various datasets from different cities while employing transfer learning to adapt the global model to the specific characteristics of individual smartphones and cities. FOGFLEET relies on an architecture that integrates edge, fog, and cloud layers, with dedicated fog nodes for each city to simplify cross-silo federated learning. Experimental results demonstrate the effectiveness of FOGFLEET framework in higher TMD accuracy by up to 20% than its comparable centralized approach. Furthermore, it outperforms the FL solutions reported in the literature with at least an 8% increase in accuracy. In this work, we also highlight the importance of sufficient training data for distributed training and discuss the impact of smartphone sensor qualities on TMD system performance. Our work contributes to advancing TMD systems by providing an adaptive and privacypreserving solution suitable for deployment in diverse urban environments and across various geographical locations.

• Kamalian, Mahdieh; Taherkordi, Amirhosein & Ferreira, Paulo (2024). TELEGAIT: Transfer Learning on Fog for Generalizable and Real-Time Transport Mode Detection. Proceedings of the IEEE. ISSN 0018-9219. s. 196–203. doi: 10.1109/FMEC62297.2024.10710203. Fulltekst i vitenarkiv Vis sammendrag

  The knowledge of users’ transport modes, e.g., cars or buses, enables innovative transportation services, such as automatic ticketing and navigation. A promising approach for transport mode detection (TMD) is analyzing smartphone sensor data to infer transport modes using machine learning (ML) techniques. However, in existing work, the accuracy of MLbased approaches for a given city is highly dependent on whether the target ML model is trained with the data collected in that specific city by various users, smartphones, etc. To address this limitation, we need a comprehensive ML-based TMD system whose accuracy is not significantly affected by the geographical location, user variation, smartphone type, or smartphone position (e.g., in a bag or pocket)—generalizable TMD. In this paper, we propose an approach, called TELeGAIT, for generalizable TMD which is based on the idea of fine-tuning a base TMD ML model for the target city. Specifically, built on a base ML model trained with the aggregated data of some different cities, we fine-tune the base model for a new city by employing Transfer Learning (TL) and using the local data of that city. Moreover, some transportation services require low-latency TMD such as automatic ticketing. To this end, TeLeGait relies on a fog-based architecture, where the fine-tuning process is offloaded to a fog node in each city, and the TMD results are also stored on the fog node. TeLeGait utilizes a Multilayer Perceptron (MLP) model trained using two approaches: segment-based and raw sensor data. Our analysis shows that the raw data-based approach is more effective for neural networks, achieving 94.42%

• Tavakolisomeh, Sanaz; Shimchenko, Marina; Ôsterlund, Erik; Bruno, Rodrigo; Ferreira, Paulo & Wrigstad, Tobias (2023). Heap Size Adjustment with CPU Control, MPLR '23: 20th ACM SIGPLAN International Conference on Managed Programming Languages and Runtimes Cascais Portugal 22 October 2023. Association for Computing Machinery (ACM). ISSN 9798400703805. s. 114–128. doi: 10.1145/3617651.3622988. Fulltekst i vitenarkiv Vis sammendrag

  This paper explores automatic heap sizing where developers let the frequency of GC expressed as a target overhead of the application's CPU utilisation, control the size of the heap, as opposed to the other way around. Given enough headroom and spare CPU, a concurrent garbage collector should be able to keep up with the application's allocation rate, and neither the frequency nor duration of GC should impact throughput and latency. Because of the inverse relationship between time spent performing garbage collection and the minimal size of the heap, this enables trading memory for computation and conversely, neutral to an application's performance. We describe our proposal for automatically adjusting the size of a program's heap based on the CPU overhead of GC. We show how our idea can be relatively easily integrated into ZGC, a concurrent collector in OpenJDK, and study the impact of our approach on memory requirements, throughput, latency, and energy.

• Vijouyeh, Lyla Naghipour; Bruno, Rodrigo & Ferreira, Paulo (2023). EdgeEmu - Emulator for Android Edge Devices. Lecture Notes in Computer Science (LNCS). ISSN 0302-9743. 13909, s. 110–127. doi: 10.1007/978-3-031-35260-7_7. Fulltekst i vitenarkiv
• Tavakolisomeh, Sanaz; Bruno, Rodrigo & Ferreira, Paulo (2023). BestGC: An Automatic GC Selector. IEEE Access. 11, s. 72357–72373. doi: 10.1109/ACCESS.2023.3294398. Fulltekst i vitenarkiv Vis sammendrag

  Garbage collection algorithms are widely used in programming languages like Java. However, selecting the most suitable garbage collection (GC) algorithm for an application is a complex task since they behave differently regarding crucial performance metrics such as garbage collection pause time, application throughput, and memory usage. This challenge is particularly more complicated as there is currently no available tool to assist users/developers in this critical decision-making process. In this paper, we address this pressing need by conducting an extensive evaluation of four widely used GCs (G1, Parallel, Shenandoah, and ZGC) in OpenJDK, considering application throughput, GC pause time, and various heap sizes. Building upon this evaluation, we present BestGC, a novel system that suggests the most suitable GC solution based on user-defined performance goals in terms of application throughput and GC pause time. Our evaluation of BestGC using multiple workloads demonstrates its effectiveness in suggesting the most suitable GC category (concurrent or generational/non-fully concurrent GC) in approximately 86% of the experiments on average. Additionally, BestGC accurately identifies the best GC in approximately 52% of the cases on average. Even in situations where BestGC failed to suggest the exact best GC or GC category, the suggested GC still outperforms the default GC (G1) in the JDK, exhibiting an average improvement of 1.75%. Notably, BestGC is designed to be easily extensible, facilitating its compatibility with other JDK versions, as well as new GCs and heap sizes. By addressing the lack of a practical tool to aid in GC selection, our research makes a significant contribution to the field of performance optimization in Java applications.

• Kamalian, Mahdieh; Ferreira, Paulo & Jul, Eric Bartley (2022). A survey on local transport mode detection on the edge of the network. Applied Intelligence - The International Journal of Research on Intelligent Systems for Real Life Complex Problems. ISSN 0924-669X. 52(14), s. 16021–16050. doi: 10.1007/s10489-022-03214-y. Fulltekst i vitenarkiv Vis sammendrag

  We present a survey of smartphone-based Transport Mode Detection (TMD). We categorize TMD solutions into local and remote; the first ones are addressed in this article. A local approach performs the following steps in the smartphone (and not in some faraway cloud servers): 1) data collection or sensing, 2) preprocessing, 3) feature extraction, and 4) classification (with a previous training phase). A local TMD approach outperforms a remote approach due to less delay, improved privacy, no need for Internet connection, better or equal accuracy and smaller data size. Therefore, we present local TMD solutions taking into account the above mentioned four steps and analyze them according to the most relevant requirements: accuracy, delay, resources consumption and generalization. To achieve the highest accuracy (100%), studies used a different combination of sensors, features and Machine Learning (ML) algorithms. The results suggest that accelerometer and GPS (Global Position System) are the most useful sensors for data collection. Discriminative ML algorithms, such as random forest, outperform the other algorithms for classification. Some solutions improved the delay of the proposed system by using a small window size and a local approach. A few studies could improve battery usage of their system by utilizing low battery-consuming sensors (e.g., accelerometer) and low sampling rate (e.g., 10Hz). CPU usage is primarily dependent on data collection, while memory usage is related to the features and complexity of the ML algorithm. Finally, the generalization requirement is met in studies that consider user, location and position independency into account.

• Helmy, Maged; Truong, Trung Tuyen; Dykyy, Anastasiya; Ferreira, Paulo & Jul, Eric Bartley (2022). CapillaryNet: An automated system to quantify skin capillary density and red blood cell velocity from handheld vital microscopy. Artificial Intelligence in Medicine. ISSN 0933-3657. 127. doi: 10.1016/j.artmed.2022.102287. Fulltekst i vitenarkiv Vis sammendrag

  Capillaries are the smallest vessels in the body which are responsible for delivering oxygen and nutrients to surrounding cells. Various life-threatening diseases are known to alter the density of healthy capillaries and the flow velocity of erythrocytes within the capillaries. In previous studies, capillary density and flow velocity were manually assessed by trained specialists. However, manual analysis of a standard 20-s microvascular video requires 20 min on average and necessitates extensive training. Thus, manual analysis has been reported to hinder the application of microvascular microscopy in a clinical environment. To address this problem, this paper presents a fully automated state-of-the-art system to quantify skin nutritive capillary density and red blood cell velocity captured by handheld-based microscopy videos. The proposed method combines the speed of traditional computer vision algorithms with the accuracy of convolutional neural networks to enable clinical capillary analysis. The results show that the proposed system fully automates capillary detection with an accuracy exceeding that of trained analysts and measures several novel microvascular parameters that had eluded quantification thus far, namely, capillary hematocrit and intracapillary flow velocity heterogeneity. The proposed end-to-end system, named CapillaryNet, can detect capillaries at ~0.9 s per frame with ~93% accuracy. The system is currently used as a clinical research product in a larger e-health application to analyse capillary data captured from patients suffering from COVID-19, pancreatitis, and acute heart diseases. CapillaryNet narrows the gap between the analysis of microcirculation images in a clinical environment and state-of-the-art systems.

• Helmy, Maged; Truong, Trung Tuyen; Jul, Eric Bartley & Ferreira, Paulo (2022). Deep Learning and Computer Vision Techniques for Microcirculation Analysis: A Review. Patterns. 4(1). doi: 10.1016/j.patter.2022.100641. Fulltekst i vitenarkiv Vis sammendrag

  The analysis of microcirculation images has the potential to reveal early signs of life-threatening diseases such as sepsis. Quantifying the capillary density and the capillary distribution in microcirculation images can be used as a biological marker to assist critically ill patients. The quantification of these biological markers is labor intensive, time consuming, and subject to interobserver variability. Several computer vision techniques with varying performance can be used to automate the analysis of these microcirculation images in light of the stated challenges. In this paper, we present a survey of over 50 research papers and present the most relevant and promising computer vision algorithms to automate the analysis of microcirculation images. Furthermore, we present a survey of the methods currently used by other researchers to automate the analysis of microcirculation images. This survey is of high clinical relevance because it acts as a guidebook of techniques for other researchers to develop their microcirculation analysis systems and algorithms.

• Tavakolisomeh, Sanaz; Bruno, Rodrigo & Ferreira, Paulo (2021). Selecting a GC for Java Applications. Norsk Informatikkonferanse (NIK). Fulltekst i vitenarkiv
• Matre, Dagfinn; Christensen, Jan Olav; Mork, Paul Jarle; Ferreira, Paulo; Sand, Trond & Nilsen, Kristian Bernhard (2021). Shift work, inflammation and musculoskeletal pain. The HUNT Study. Occupational Medicine. ISSN 0962-7480. 71(9), s. 422–427. doi: 10.1093/occmed/kqab133. Fulltekst i vitenarkiv Vis sammendrag

  Background: Studies have indicated that shift work, in particular night work, is associated with chronic musculoskeletal pain but the mechanisms are unclear. It has been suggested that sleep disturbance, a common complaint among shift and night workers, may induce low-grade inflammation as well as heightened pain sensitivity. Aims: Firstly, this study was aimed to examine the cross-sectional associations between shift work, C-reactive protein (CRP) level and chronic musculoskeletal pain, and secondly, to analyse CRP as a mediator between shift work and chronic musculoskeletal pain. Methods: The study included 23 223 vocationally active women and men who participated in the HUNT4 Survey of the Trøndelag Health Study (HUNT). Information was collected by questionnaires, interviews, biological samples and clinical examination. Results: Regression analyses adjusted for sex, age and education revealed significant associations between shift work and odds of any chronic musculoskeletal pain (odd ratio [OR] 1.11, 95% confidence interval [CI] 1.04-1.19), between shift work and CRP level (OR 1.09, 95% CI 1.03-1.16) and between CRP level 3.00-10 mg/L and any chronic musculoskeletal pain (OR 1.38, 95% CI 1.27-1.51). Shift work and CRP were also associated with number of chronic pain sites. Mediation analysis indicated that shift work was indirectly associated with any chronic musculoskeletal pain through CRP (OR 1.03, 95% CI 1.01-1.06). Conclusions: The results support the hypothesis that shift work is associated with chronic musculoskeletal pain, and that systemic inflammation may be a biological mechanism linking shift work to chronic pain.

• Ferreira, Paulo; Zavgorodnii, Constantin & Veiga, Luìs (2020). edgeTrans - Edge transport mode detection. Pervasive and Mobile Computing. ISSN 1574-1192. 69. doi: 10.1016/j.pmcj.2020.101268. Fulltekst i vitenarkiv Vis sammendrag

  Automatic human activity detection, mainly transport-wise, is very relevant for architects and urban planners (among many others) when designing cities, roads, public transportation systems, etc. Such detection allows to better plan our cities and has been made possible with the widespread use of smartphones carrying several different sensors. We developed edgeTrans, a system based on a smartphone application (app, for short), a database, and a server. The database stores the trips that were done, the server runs a machine learning algorithm that generates a model (i.e., a classifier) which is then integrated into the edgeTrans app. This app, after being installed, when running in a smartphone indicates the transport mode that is being used without requiring a network connection; it can now be downloaded from the Android Play Store or from the iOS iTunes (the app is called Woorti). The results obtained in a real-world setting are very encouraging taking into account the requirements (e.g., accuracy, and low cost).

• Bruno, Rodrigo; Patricio, Duarte; Simão, José; Veiga, Luìs & Ferreira, Paulo (2019). Runtime Object Lifetime Profiler for Latency Sensitive Big Data Applications, EuroSys '19: Fourteenth EuroSys Conference 2019. Association for Computing Machinery (ACM). ISSN 9781450362818. doi: 10.1145/3302424.3303988. Fulltekst i vitenarkiv
• Ferreira, Paulo; Zabolotny, Andriy & Barreto, João (2019). Bicycle Mode Activity Detection with Bluetooth Low Energy Beacons. I Gkoulalas-Divanis, Aris; Marchetti, Mirco & Avresky, Dimiter R. (Red.), 2019 IEEE 18th International Symposium on Network Computing and Applications (NCA). IEEE (Institute of Electrical and Electronics Engineers). ISSN 9781728125220. doi: 10.1145/3412841.3441965. Fulltekst i vitenarkiv
• Tavakolisomeh, Sanaz; Shimchenko, Marina; Österlund, Erik; Bruno, Rodrigo; Ferreira, Paulo & Wrigstad, Tobias (2008). Heap Size Adjustment with CPU Control. I Vasconcelos, Vasco Thudichum (Red.), SPLASH '23: 2023 ACM SIGPLAN International Conference on Systems, Programming, Languages, and Applications: Software for Humanity. Association for Computing Machinery (ACM). ISSN 9798400703843. s. 2023–2023. doi: 10.1145/3617651.3622988. Fulltekst i vitenarkiv

Se alle arbeider i NVA

• Ferreira, Paulo (2018). Proceedings of the 19th International Middleware Conference. Association for Computing Machinery (ACM). ISBN 9781450357029. 299 s. Fulltekst i vitenarkiv

Se alle arbeider i NVA

• Vijouyeh, Lyla Naghipour; Veiga, Luis & Ferreira, Paulo (2025). BCProf: Battery Consumption Profiler for Android Applications. doi: https:/link.springer.com/book/10.1007/978-3-031-95728-4. Fulltekst i vitenarkiv Vis sammendrag

  The widespread use of mobile applications in daily life has raised significant concerns about their energy consumption, as it directly affects user satisfaction, their hosting device longevity, and environmental sustainability. Despite the need for energy-aware mobile application development, developers lack cost-effective, software-based, and intuitive tools to measure and analyze energy usage during the development process. This paper introduces BCProf (Battery Consumption Profiler), a novel framework designed to measure the energy consumption of Android applications at both method and application granularities. BCProf supports both Java and Kotlin languages and is implemented as a plugin for seamless integration into Android Studio. BCProf works by instrumenting the application’s source code with custom log statements that enable tracking Android API calls. During application testing, BCProf monitors the generated logs alongside hardware component usage (e.g., WiFi, GPS, camera, Bluetooth, and screen) to calculate energy consumption. By leveraging this monitored data, BCProf correlates the energy usage of Android API calls and hardware components with specific methods in the application’s source code. The tool supports both online, real-time energy analysis during application execution and offline analysis by saving energy usage data in a JSON file format. Extensive evaluations were conducted across various Android applications to validate the accuracy and versatility of BCProf. The results show that BCProf can achieve up to 98% accuracy. This makes it a reliable and adaptable solution for the energy profiling of Android mobile applications.

• Ferreira, Paulo (2023). Vil bruke KI til å analysere transportdata fra smarttelefoner – kan føre til nye billettsystemer. [Journal]. Forskningparken. Fulltekst i vitenarkiv
• Kamalian, Mahdieh & Ferreira, Paulo (2022). FogTMDetector - Fog Based Transport Mode Detection using Smartphones. Fulltekst i vitenarkiv Vis sammendrag

  A user’s transport mode (e.g., walk, car, etc.) can be detected by using a smartphone. Such devices exist in a great number with enough computation power and sensors to run a classifier (i.e., for transport mode detection). Using a smartphone in a fog environment ensures low latency, high generalization, high accuracy, and low battery consumption. We propose a fog- based real-time (at human time scale) transport mode detection, called FogTMDetector; it consists of a Random Forest classifier trained with magnetometer, accelerometer, and GPS data. The overall accuracy achieved by our system is 93% when detecting 8 different modes (i.e., stationary, walk, bicycle, car, bus, train, tram, and subway). We compared FogTMDetector with another recent system (called EdgeTrans). The comparison results suggest that our solution achieves 10% higher motorized accuracy (i.e., 94.4%) with more fine-grained motorized transport modes (i.e., subway, tram, etc.) thanks to the magnetometer sensor readings. FogTMDetector uses a low sampling rate (1Hz) for logging accelerometer and magnetometer and (every 10 seconds) for GPS to ensure low battery consumption. FogTMDetector is also generalizable as it is robust against variation of users and smartphone positions.

• Ferreira, Paulo (2018). Foreword for the 19th ACM/IFIP/USENIX Middleware Conference. I Ferreira, Paulo (Red.), Proceedings of the 19th International Middleware Conference. Association for Computing Machinery (ACM). ISSN 9781450357029. s. 1–3. Fulltekst i vitenarkiv
• Ferreira, Paulo & Bruno, Rodrigo (2018). Live Migration and Garbage Collection for Big Data Java Applications. Fulltekst i vitenarkiv

Se alle arbeider i NVA