In agent-based social simulations (ABSS), an artificial population of intelligent agents that imitate human behavior is used to investigate complex phenomena within social systems. This is particularly useful for decision makers, where ABSS can provide a sandpit for investigating the effects of policies prior to their implementation. During the Covid-19 pandemic, for instance, sophisticated models of human behavior enable the investigation of the effects different interventions can have and even allow for analyzing why a certain situation occurred or why a specific behavior can be observed. In contrast to other applications of simulation, the use for policy making significantly alters the process of model building and assessment, and requires the modelers to follow different paradigms. In this chapter, we report on a tutorial that was organized as part of the ACAI 2021 summer school on AI in Berlin, with the goal of introducing agent-based social simulation as a method for facilitating policy making. The tutorial pursued six Intended Learning Outcomes (ILOs), which are accomplished by three sessions, each of which consists of both a conceptual and a practical part. We observed that the PhD students participating in this tutorial came from a variety of different disciplines, where ABSS is mostly applied as a research method. Thus, they do often not have the possibility to discuss their approaches with ABSS experts. Tutorials like this one provide them with a valuable platform to discuss their approaches, to get feedback on their models and architectures, and to get impulses for further research.

We present an O(n² · min{m, n}) time and O(n · min{m, n}) storage algorithm to compute the minsum set of m watchmen routes given their starting points in a Stiegl polygon ― a staircase polygon where the floor solely consists of one horizontal and one vertical edge ― with n vertices.

The development in smart meter technology has made grid operations more efficient based on fine-grained electricity usage data generated at different levels of time granularity. Consequently, machine learning algorithms have benefited from these data to produce useful models for important grid operations. Although machine learning algorithms need historical data to improve predictive performance, these data are not readily available for public utilization due to privacy issues. The existing smart grid data simulation frameworks generate grid data with implicit privacy concerns since the data are simulated from a few real energy consumptions that are publicly available. This paper addresses two issues in smart grid. First, it assesses the level of privacy violation with the individual household appliances based on synthetic household aggregate loads consumption. Second, based on the findings, it proposes two privacy-preserving mechanisms to reduce this risk. Three inference attacks are simulated and the results obtained confirm the efficacy of the proposed privacy-preserving mechanisms.

Deep reinforcement learning has advanced signifi-cantly in recent years, and it is now used in embedded systems in addition to simulators and games. Reinforcement Learning (RL) algorithms are currently being used to enhance device operation so that they can learn on their own and offer clients better services. It has recently been studied in a variety of industrial applications. However, reinforcement learning, especially when controlling a large number of agents in an industrial environment, has been demonstrated to be unstable and unable to adapt to realistic situations when used in a real-world setting. To address this problem, the goal of this study is to enable multiple reinforcement learning agents to independently learn control policies on their own in dynamic industrial contexts. In order to solve the problem, we propose a dynamic multi-agent reinforcement learning (dynamic multi-RL) method along with adaptive exploration (AE) and vector-based action selection (VAS) techniques for accelerating model convergence and adapting to a complex industrial environment. The proposed algorithm is tested for validation in emergency situations within the telecommunications industry. In such circumstances, three unmanned aerial vehicles (UAV-BSs) are used to provide temporary coverage to mission-critical (MC) customers in disaster zones when the original serving base station (BS) is destroyed by natural disasters. The algorithm directs the participating agents automatically to enhance service quality. Our findings demonstrate that the proposed dynamic multi-RL algorithm can proficiently manage the learning of multiple agents and adjust to dynamic industrial environments. Additionally, it enhances learning speed and improves the quality of service.

This paper proposes a novel agglomerated privacy-preservation model integrated with data mining and evolutionary Genetic Algorithm (GA). Privacy-pReservIng with Minimum Epsilon (PRIMϵ) delivers minimum privacy budget (ϵ) value to protect personal or sensitive data during data mining and publication. In this work, the proposed Pattern identification in the Locale of Users with Mining (PLUM) algorithm, identifies frequent patterns from dataset containing users’ sensitive data. ϵ-allocation by Differential Privacy (DP) is achieved in PRIMϵ with GA PRIMϵ , yielding a quantitative measure of privacy loss (ϵ) ranging from 0.0001 to 0.045. The proposed model maintains the trade-off between privacy and data utility with an average relative error of 0.109 on numerical data and an Earth Mover’s Distance (EMD) metric in the range between [0.2,1.3] on textual data. PRIMϵ model is verified with Probabilistic Computational Tree Logic (PCTL) and proved to accept DP data only when ϵ ≤ 0.5. The work demonstrated resilience of model against background knowledge, membership inference, reconstruction, and privacy budget attack. PRIMϵ is compared with existing techniques on DP and is found to be linearly scalable with worst time complexity of O(n log n) .

In this paper an experimental system assisted by emerging digital technologies is developed, for monitoring and controlling invasive water hyacinth weed and rescue operation of freshwater lakes in Africa. The system is designed to integrate fifth generation ultra-reliable low latency communication (5G URLLC), unmanned aerial vehicles (UAV), underwater robots, smart environmental sensing with internet of things (IoT) and machine learning techniques for real time monitoring, managing, controlling and predicting the expansion of invasive water hyacinth weed. The experimental system for sensor data collection implemented on lake Tana in Ethiopia will be expanded to other fresh water lakes of Africa affected by water hyacinth weed. System modeling and data analytics based on sensor data will be performed to generate decision inference for controlling the growth of water hyacinth in the water bodies of the lake. Environmental data collection from other local sources will be integrated with sensor data for further system modeling and critical action analysis and implementation using machine learning algorithms to remove the main causes for the rapid expansion of water hyacinth throughout the lake.

Data security has remained a challenging problem in cloud storage, especially in multiowner data sharing scenarios. As one of the most effective solutions for secure data sharing, multikeyword ranked searchable encryption (MRSE) has been widely used. However, most of the existing MRSE schemes have some shortcomings in multiowner data sharing, such as index trees generated by data owners, relevance scores in plaintext form, and lack of verification function. In this article, we propose a verifiable and efficient secure sharing scheme in multiowner multiuser settings, where the index tree is generated by the trusted authority. To achieve verifiable functionality, the blockchain-based smart contract is adopted to execute the search algorithm. Based on a distributed two-trapdoor public-key cryptosystem, the data uploaded and used are in ciphertext form, and the proposed algorithms are secure in our scheme. For improving efficiency, the encrypted data are aggregated according to the category and the Category ID-based index tree is generated. Extensive experiments are conducted to demonstrate that it can reduce the time cost of index construction by 75% and the time cost of search by 53%, approximately. Moreover, multithreaded optimization is introduced in our scheme, which can reduce the time cost of index construction by 76% and the time cost of search by 67%, approximately (with 16 threads).

Our dependency on Global Navigation Satellite System (GNSS) for getting directions, tracking items, locating friends, or getting maps of the world has increased tremendously over the last decade. However, as soon as we enter a building, the signal strength of the satellites is too low, and we need to resort to other technologies to achieve the same goals. An Indoor Positioning System (IPS) may utilize a wide range of methods for positioning a device, such as fingerprinting, multilateration, or sensor fusion, while using one or several radio technologies to measure Received Signal Strength (RSS) or Time of Arrival(ToA). Sensor fusion is an efficient approach where an Inertial Measurement Unit (IMU) is combined with, e.g., RSS measurements converted to distances. But this approach has significant drawbacks in areas where, e.g., walls or large objects obstruct the signal path, which introduces bias in the distance estimates. This paper addresses the bias caused by signal path obstruction by compensating the measured RSS with localized RSS attenuation adjustments and thereby increasing the accuracy of the sensor fusion model significantly. We also show that a system can learn the compensation parameters over time, reducing the installationefforts and achieving higher accuracy than a fingerprinting-based system.

Our phones have many uses for positioning technologies, such as navigation, LocationBased Services (LBS), emergency positioning, fitness applications, and advertising. We trust our phones and wearables to be location-aware. However, as soon as we enter a building, we can no longer use GPS signals, as their already weak signals are well below the background noise of the environment. This requires us to develop alternatives, such as installing active radio beacons, using existing radio infrastructure, applying environmental sensing based on barometric pressure and magnetic fields, or utilizing Inertial Measurement Units (IMUs) to estimate the user location. This licentiate thesis aims to evaluate beacon-based indoor positioning, where we assume installing a set of small battery-powered Bluetooth low-energy (BLE) beacons are possible. In particular, the thesis addresses essential factors such as installation effort, accuracy, the privacy aspects of an Indoor Positioning System(IPS), and mitigation of accuracy issues related to radio signal shadowing in complex indoor environments. The goal is to solve some obstacles to the widespread adoption of indoor positioning solutions.

The demand for the accurate and timely identification of melanoma as a major skin cancer type is increasing daily. Due to the advent of modern tools and computer vision techniques, it has become easier to perform analysis. Skin cancer classification and segmentation techniques require clear lesions segregated from the background for efficient results. Many studies resolve the matter partly. However, there exists plenty of room for new research in this field. Recently, many algorithms have been presented to preprocess skin lesions, aiding the segmentation algorithms to generate efficient outcomes. Nature-inspired algorithms and metaheuristics help to estimate the optimal parameter set in the search space. This research article proposes a hybrid metaheuristic preprocessor, BA-ABC, to improve the quality of images by enhancing their contrast and preserving the brightness. The statistical transformation function, which helps to improve the contrast, is based on a parameter set estimated through the proposed hybrid metaheuristic model for every image in the dataset. For experimentation purposes, we have utilised three publicly available datasets, ISIC-2016, 2017 and 2018. The efficacy of the presented model is validated through some state-of-the-art segmentation algorithms. The visual outcomes of the boundary estimation algorithms and performance matrix validate that the proposed model performs well. The proposed model improves the dice coefficient to 94.6% in the results.

On-demand transport services are often envisioned as stand-alone modes or as a replacement for conventional public transport modes. This leads to a comparison of service efficiencies, or direct competition for passengers between them. The results of this work point to the positive effects of the inclusion of DRT into the public transport network. We simulate a day of operation of a DRT service in a rural area and demonstrate that a DRT system that focuses on increasing accessibility for travellers with poor public transport access can be quite efficient, especially for reducing environmental impact. We show that DRT, while it produces more vehicle kilometres than private cars would inside the DRT operating zone, can help to reduce the vehicle kilometres travelled for long-distance trips. The results of this study indicate the need for a more systemic evaluation of the impact of the new mobility modes.

Public transport is an efficient way to transport large volumes of travellers. However, there are systemic issues that make it hard for conventional public transport to provide efficient service on finer levels, like first- and last-mile problems or low-demand areas. One of the potential solutions that has been getting a lot of attention recently in research and real practice is Demand-Responsive Transport(DRT). The main difference between demand-responsive services and conventional public transport is the need for explicit requests for a trip from the travellers. The service then adapts the routes of the vehicles to satisfy the requests as efficiently as possible. One of the aims of such transport services is to combine the flexibility and accessibility of travel modes like taxis and private cars with the efficiency of buses achieved through ride-sharing.DRT has the potential to improve public transport in, for example, low population density areas or for people with mobility limitations who could request a trip directly to a home door. Historically DRT has been extensively used for special transportation while the recent trend in research and practice explores the possibility of using this service type for the general population.The history of DRT shows a large degree of discontinued trials and services together with low utilisation of vehicles and limited efficiency levels. In practice, this leads to measures restricting the trip destination, times when service is available, or eligibility to use the service at all in case of special transport DRT. Due to the limited use of DRT services, there is little data collected on the efficiency of the service and transport agencies exploring the possibility of introducing this new service type face difficulties in estimating its potential.The main goal of this thesis is to contribute towards developing a decisionsupport method for transport analysts, planners, or decision-makers who want to evaluate the systemic effect of a DRT service such as costs, emissions and effecton society. Decision-makers should be able to evaluate and compare a large variety of DRT design choices like booking time restrictions, vehicle fleet type, target trip quality level, or stop allocation pattern. Using a design science, we develop a simulation approach which is evaluated with two simulation experiments. The simulation experiments themselves provide valuable insight into the potential of DRT services, explore the niche where DRT could provide the most benefits and advocate taking into account the sustainability perspective for a comprehensive comparison of transport modes.

The findings from the simulation experiments indicate that DRT, even in its extreme forms like fully autonomous shared taxis, does not show the level of efficiency that could result in a revolution in transportation — it is hard to compete inefficiency with conventional public transport in urban zones. However, in scenarios with lower demand levels, it could be more efficient to replace conventional buses with a DRT service when considering costs and emissions. We also show that, when integrated with conventional public transport, DRT could help alleviate the last-mile problem by improving accessibility to long-distance lines. Additionally, if car users are attracted to public transport with the help of DRT, there is a potential to significantly reduce the total level of emissions.

The simulation results indicate that the proposed simulation method can be applied for the evaluation of DRT. The implementation of trip planning combining DRT and conventional public transport is a major contribution of this thesis. We show that the integration between services may be important for the efficiency of the service, especially when considering the sustainability aspects.

Finally, this thesis indicates the direction for further research. The proposed simulation approach is suitable for the estimation of the potential of DRT but lacks the ability to make a prediction of the demand for DRT. Integration of a realistic mode choice model and day-to-day simulations are important for making predictions. We also note the complexity of the DRT routing for large-scale problems which prohibits a realistic estimation with simulation and the efficient operation of the service.

Sverige ses som ett av de mest forskningsdrivna och utbildade länderna i världen. Forskarutbildningen anses därmed vara en av Sveriges viktigaste delar av den högre utbildningen. Detta ställningstagande reflekterar över forskarutbildningens process och produkt i Datavetenskap och programvarusystem i Sverige. Artikeln ger en översikt över forskarutbildningen i Sverige, följt av en praktisk demonstration av hur handledare och doktorander kan använda konstruktiv anpassning för att uppnå lärandemålen för forskarutbildningen med hjälp av inlärningsaktiviteter och bedömningsmetoder för att utvärdera lärandeaktiviteterna och i förlängningen lärandemålen. 

Deep reinforcement learning has made significant development in recent years, and it is currently applied not only in simulators and games but also in embedded systems. However, when implemented in a real-world context, reinforcement learning is frequently shown to be unstable and incapable of adapting to realistic situations, particularly when directing a large number of agents. In this paper, we develop a decentralized architecture for reinforcement learning to allow multiple agents to learn optimal control policies on their own devices of the same kind but in varied environments. For such multiple agents, the traditional centralized learning algorithm usually requires a costly or time-consuming effort to develop the best-regulating policy and is incapable of scaling to a large-scale system. To address this issue, we propose a decentralized reinforcement learning algorithm (DecRL) and information exchange scheme for each individual device, in which each agent shares the individual learning experience and information with other agents based on local model training. We incorporate the algorithm into each agent in the proposed collaborative architecture and validate it in the telecommunication domain under emergency conditions, in which a macro base station (BS) is broken due to a natural disaster, and three unmanned aerial vehicles carrying BSs (UAV-BSs) are deployed to provide temporary coverage for missioncritical (MC) users in the disaster area. Based on the findings, we show that the proposed decentralized reinforcement learning algorithm can successfully support multi-agent learning, while the learning speed and service quality can be further enhanced.

Interactive machine learning (ML) adds a human-in-the-loop aspect to a ML system. Even though the input from human users to the system is a central part of the concept, the uncertainty caused by the human feedback is often not considered in interactive ML. The assumption that the human user is expected to always provide correct feedback, typically does not hold in real-world scenarios. This is especially important for when the cognitive workload of the human is high, for instance in online learning from streaming data where there are time constraints for providing the feedback. We present experiments of interactive online ML with human participants, and compare the results to simulated experiments where humans are always correct. We found combining the two interactive learning paradigms, active learning and machine teaching, resulted in better performance compared to machine teaching alone. The results also showed an increased discrepancy between the experiments with human participants and the simulated experiments when the cognitive workload was increased. The findings suggest the importance of taking uncertainty caused by human factors into consideration in interactive ML, especially in situations which requires a high cognitive workload for the human.

This paper concerns the novel concept of an Interactive Dynamic Intelligent Virtual Sensor (IDIVS), which extends virtual/soft sensors towards making use of user input through interactive learning (IML) and transfer learning. In research, many studies can be found on using machine learning in this domain, but not much on using IML. This paper contributes by highlighting how this can be done and the associated positive potential effects and challenges. An IDIVS provides a sensor-like output and achieves the output through the data fusion of sensor values or from the output values of other IDIVSs. We focus on settings where people are present in different roles: from basic service users in the environment being sensed to interactive service users supporting the learning of the IDIVS, as well as configurators of the IDIVS and explicit IDIVS teachers. The IDIVS aims at managing situations where sensors may disappear and reappear and be of heterogeneous types. We refer to and recap the major findings from related experiments and validation in complementing work. Further, we point at several application areas: smart building, smart mobility, smart learning, and smart health. The information properties and capabilities needed in the IDIVS, with extensions towards information security, are introduced and discussed.

Cooperative work differs depending on contexts and tasks, whether co-located, synchronous, or distributed in time and space. New technology allows new opportunities to support cooperation. A central aspect of cooperation is the relation to individual work; when co-located, people enter and exit cooperation seamlessly. This dissertation explores how technology, situation, and context interplay in various forms of cooperation. It addresses two research questions: (1) How do people get engaged in cooperative work? and (2) How can engagement in distributed cooperative work be supported?

The work focuses on ethnographic empirical studies that analyse the interaction between humans and technology across various domains. Workplace studies have been conducted in different fields. Emergency service work, truck driver's work, building maintenance workers, and visitor's technology use at a music festival. The workplace studies in the dissertation imply that field studies are conducted to document and analyse how people use technology and how this use takes place. Common to all studies is the work about activities distributed in time and space.

These research findings inform the development of new perspectives, concepts, and design challenges for distributed collaboration. The dissertation discusses two primary ways to engage in cooperative work are identified: requesting and choosing to engage through shared materials and artefacts support awareness and enable cooperative work. The results identify four factors to facilitate engagement in remote cooperative environments: supporting requests and choices to engage, providing opportunities to use artefacts, promoting shareability, and incorporating awareness technology.

The dissertation contributes new insights into the interplay between technology, situation, and context in cooperation. Providing design insights for distributed collaboration, and the exploration of design concepts and analysis models. The contributions emphasize the dynamic nature of collaboration and the importance of understanding the relationship between individual and cooperative work to support distributed and remote collaboration effectively.

The use of artificial intelligence (AI) is increasing in our everyday applications. One emerging field within AI is image recognition. Research that has been devoted to predicting fires involves predicting its behaviour. That is, how the fire will spread based on environmental key factors such as moisture, weather condition, and human presence. The result of correctly predicting fire spread can help firefighters to minimise the damage, deciding on possible actions, as well as allocating personnel effectively in potentially fire prone areas to extinguish fires quickly. Using neural networks (NN) for active fire detection has proven to be exceptional in classifying smoke and being able to separate it from similar patterns such as clouds, ground, dust, and ocean. Recent advances in fire detection using NN has proved that aerial imagery including drones as well as satellites has provided great results in detecting and classifying fires. These systems are computationally heavy and require a tremendous amount of data. A NN model is inextricably linked to the dataset on which it is trained. The cornerstone of this study is based on the data dependencieds of these models. The model herein is trained on two separate datasets and tested on three dataset in total in order to investigate the data dependency. When validating the model on their own datasets the model reached an accuracy of 92% respectively 99%. In comparison to previous work where an accuracy of 94% was reached. During evaluation of separate datasets, the model performed around the 60% range in 5 out of 6 cases, with the outlier of 29% in one of the cases. 

Artificiell intelligens och i synnerhet Maskininlärning (ML) påverkar i hög grad människors liv genom de kan skapa monetärt värde från data. Denna produktifiering av insamlad data påverkar på många sätt våra liv, från val av partner till att rekommendera nästa produkt att konsumera. ML-baserade system fungerar väl i denna roll eftersom de kan förutsäga människors beteende baserat på genomsnittliga prestandamått, men deras användbarhet är mer begränsad i situationer där det är viktigt med transparens visavi de kunskapsrepresentationer ett enskilt beslut baseras på.

 Målet med detta arbete är att kombinera människors och maskiners styrkor via en tydlig maktrelation där en slutanvändare har kommandot. Denna maktrelation bygger på användning av ML-system som är transparenta med bakomliggande orsaker för ett föreslaget beslut. Artificiella neurala nätverk är ett intressant val av ML-teknik för denna uppgift eftersom de kan bygga interna kunskapsrepresentationer från rå data och därför tränas utan specialiserad ML kunskap. Detta innebär att ett neuralt nätverk kan tränas genom att exponeras för data från en måldomän och i denna process internalisera relevanta kunskapsrepresentationer. Därefter kan nätet presentera kontextuella förslag på beslut baserat på dessa representationer. I icke-statiska situationer behöver det fragment av den verkliga världen som internaliseras i ML-systemet kontextualiseras av en människa för att systemet skall vara användbart och tillförlitligt.

 I detta arbete utforskas det ovan beskrivna området via en övergripande forskningsfråga: Vilka utmaningar och möjligheter kan uppstå när en slutanvändare använder neurala nätverk som stöd för enstaka beslut i ett väldefinierat sammanhang?

 För att besvara forskningsfrågan ovan används metodologin forskning genom design, detta på grund av att den valda metodologin matchar öppenheten i forskningsfrågan. Genom sex designexperiment utforskas utmaningar och möjligheter i situationer där enskilda kontextuella beslut är viktiga. De initiala designexperimenten fokuserar främst på möjligheter i situationer där neurala nätverk presterar i paritet med människors kognitiva förmågor och de senare experimenten utforskar utmaningar i situationer där neurala nätverk överträffar människans kognitiva förmågor.  Den andra delen fokuserar främst på metoder som syftar till att förklara beslut föreslagna av det neurala nätverket.

 Detta arbete bidrar till existerande kunskap på tre sätt: (1) utforskande av lärande relaterat till neurala nätverk med målet att presentera en terminologi användbar för kontextuellt beslutsfattande understött av ML-system, den framtagna terminologin inkluderar generativa begrepp som: sann-i-relation-till-domänen, koncept, utanför-distributionen och generalisering, (2) ett antal designriktlinjer, (3) behovet av att justera interna kunskapsrepresentationer i neurala nätverk så att de överensstämmer med koncept vilket skulle kunna medföra att neurala nätverk kan producera förklaringsbara beslut. Jag föreslår även att en framkomlig forskningsstrategi är att träna neurala nätverk med utgångspunkt från grundläggande koncept, som former och färger. Denna strategi innebär att nätverken kan generalisera utifrån dessa generella koncept i olika domäner. Den föreslagna forskningsriktning syftar till att producera mer komplexa förklaringar från neurala nätverk baserat på grundläggande generaliserbara koncept.

Neuropathic pain is a difficult condition to treat and would require reliable biomarkers to personalise and optimise treatments. To date, pain levels are mostly measured with subjective scales, but research has shown that electroencephalography (EEG) and heart rate variability (HRV) can be linked to those levels. Internet of Things technology could allow embedding EEG and HRV in easy-to-use systems that patients can use at home in their daily life. We have developed a system for home monitoring that includes a portable EEG device, a tablet application to guide patients through imaginary motor tasks while recording EEG, a wearable HRV sensor and a mobile phone app to report pain levels. We are using this system in a clinical study involving 15 spinal cord injury patients for one month. Preliminary results show that relevant data are being collected, with inter and intra-patients variability for both HRV and pain levels, and that the mobile phone app is perceived as usable, of good quality and useful. However, because of its complexity, the system requires some effort from patients, is sometimes unreliable and the collected EEG signals are not always of the desired quality.

Asthma monitoring is an important aspect of patient self-management. However, due to its repetitive nature, patients can find long-term monitoring tedious. Mobile health can provide an avenue to monitor asthma without needing high levels of active engagement, and instead rely on passive monitoring. In our recent AAMOS-00 study, we collected mobile health data over six months from 22 asthma patients using passive and active monitoring technology, including smartwatch, peak flow measurements, and daily asthma diaries.

Compliance to smartwatch monitoring was found to lie between the compliance to complete daily asthma diaries and measuring daily peak flow. However, some study participants faced technical issues with the devices which could have affected the relative compliance of the monitoring tasks.

Moreover, as evidenced by standard usability questionnaires, we found that the AAMOS-00 study’s data collection system was similar in quality to other studies and published apps.

The currently dominating artificial intelligence and machine learning technology, neural networks, builds on inductive statistical learning processes. Being void of knowledge that can be used deductively these systems cannot distinguish exemplars part of the target domain from those not part of it. This ability is critical when the aim is to build human trust in real-world settings and essential to avoid usage in domains wherein a system cannot be trusted. In the work presented here, we conduct two qualitative contextual user studies and one controlled experiment to uncover research paths and design openings for the sought distinction. Through our experiments, we find a need to refocus from average case metrics and benchmarking datasets toward systems that can be falsified. The work uncovers and lays bare the need to incorporate and internalise a domain ontology in the systems and/or present evidence for a decision in a fashion that allows a human to use our unique knowledge and reasoning capability. Additional material and code to reproduce our experiments can be found at https://github.com/k3larra/ood.

In configuration interaction (CI) calculations the atomic wave functions are given as expansions over configuration state functions (CSFs) built on relativistic one-electron orbitals. The expansion coefficients of the configuration state functions are obtained by constructing and diagonalizing the Hamiltonian matrix. Here we show how a regrouping of the configuration state functions and the introduction of configuration state function generators (CSFGs) allow for a substantial reduction of the computational load in relativistic CI calculations. The computational methodology based on configuration state function generators, recently implemented in the General Relativistic Atomic Structure package (Grasp2018, Froese Fischer et al. (2019) [16]), is applied to a number of atomic systems and correlation models with increasing sets of one-electron orbitals. We demonstrate a reduction of the CPU time with factors between 10 and 14 for the largest CI calculations. The inclusion of the Breit interaction into the calculations is time consuming. By applying restrictions on the Breit integrals we show that it is possible to further reduce the CPU times with factors between 2 and 3, with negligible changes to the computed excitation energies. We also demonstrate that the introduction of configuration state function generators allows for efficient a priori condensation techniques, with reductions of the expansions sizes with factors between 1.5 and 2.5 and the CPU time with factors between 2.5 and 4.5, again with negligible changes to the excitation energies. In total we demonstrate reductions of the CPU time with factors up to 68 for CI calculations based on configuration state function generators, restrictions on the Breit integrals and with a priori condensed expansions compared to ordinary CI calculations without restrictions on the Breit integrals and with full expansions. Further perspectives of the new methodology based on configuration state function generators are given.

Artificial intelligence (AI) and the use of machine learning (ML) and deep learning (DL) technologies are becoming increasingly popular in companies. These technologies enable companies to leverage big quantities of data to improve system performance and accelerate business development. However, despite the appeal of ML/DL, there is a lack of systematic and structured methods and processes to help data scientists and other company roles and functions to develop, deploy and evolve models. In this paper, based on multi-case study research in six companies, we explore practices and challenges practitioners experience in developing ML/DL models as part of large software-intensive embedded systems. Based on our empirical findings, we derive a conceptual framework in which we identify three high-level activities that companies perform in parallel with the development, deployment and evolution of models. Within this framework, we outline activities, iterations and triggers that optimize model design as well as roles and company functions. In this way, we provide practitioners with a blueprint for effectively integrating ML/DL model development into the business to achieve better results than other (algorithmic) approaches. In addition, we show how this framework helps companies solve the challenges we have identified and discuss checkpoints for terminating the business case.

Reinforcement learning, particularly deep reinforcement learning, has made remarkable progress in recent years and is now used not only in simulators and games but is also making its way into embedded systems as another software-intensive domain. However, when implemented in a real-world context, reinforcement learning is typically shown to be fragile and incapable of adapting to dynamic environments. In this paper, we provide a novel dynamic reinforcement learning algorithm for adapting to complex industrial situations. We apply and validate our approach using a telecommunications use case. The proposed algorithm can dynamically adjust the position and antenna tilt of a drone-based base station to maintain reliable wireless connectivity for mission-critical users. When compared to traditional reinforcement learning approaches, the dynamic reinforcement learning algorithm improves the overall service performance of a drone-based base station by roughly 20%. Our results demonstrate that the algorithm can quickly evolve and continuously adapt to the complex dynamic industrial environment.

Rapid advancements and the escalating necessity of autonomous algorithms in medical imaging require efficient models to accomplish tasks such as segmentation and classification. However, there exists a significant dependency on the image quality of datasets when using these models. Appreciable improvements to enhance datasets for efficient image analysis have been noted in the past. In addition, deep learning and machine learning are vastly employed in this field. However, even after the advent of these advanced techniques, a significant space exists for new research. Recent research works indicate the vast applicability of preprocessing techniques in segmentation tasks. Contrast stretching is one of the preprocessing techniques used to enhance a region of interest. We propose a novel hybrid meta-heuristic preprocessor (DE-ABC), which optimises the decision variables used in the contrast-enhancement transformation function. We validated the efficiency of the preprocessor against some state-of-the-art segmentation algorithms. Publicly available skin-lesion datasets such as 𝑃𝐻², ISIC-2016, ISIC-2017, and ISIC-2018 were employed. We used Jaccard and the dice coefficient as performance matrices; at the maximum, the proposed model improved the dice coefficient from 93.56% to 94.09%. Cross-comparisons of segmentation results with the original datasets versus the contrast-stretched datasets validate that DE-ABC enhances the efficiency of segmentation algorithms.

As connectivity has been introduced to the car industry, automotive companies have in-use cars which are connected to the internet. A key concern in this context represents the difficulty of knowing how the connection quality changes over time and if there are associated issues. In this work we describe the use of CDR data from connected cars supplied by Volvo to build and study forecasting models that predict how relevant KPIs change over time. Our experiments show promising results for this predictive task, which can lead to improving user experience of connectivity in smart vehicles.

Accent invariance in speech recognition is a chal- lenging problem especially in the are of aviation. In this paper a speech recognition system is developed to transcribe accented speech between pilots and air traffic controllers. The system allows handling of accents in continuous speech by modelling phonemes using Hidden Markov Models (HMMs) with Gaussian mixture model (GMM) probability density functions for each state. These phonemes are used to build word models of the NATO phonetic alphabet as well as the numerals 0 to 9 with transcriptions obtained from the Carnegie Mellon University (CMU) pronouncing dictionary. Mel-Frequency Cepstral Co-efficients (MFCC) with delta and delta-delta coefficients are used for the feature extraction process. Amplitude normalisation and covariance scaling is implemented to improve recognition accuracy. A word error rate (WER) of 2% for seen speakers and 22% for unseen speakers is obtained.

This paper explores the problem of determining the time of an analogue wristwatch by developing two systems and conducting a comparative study. The first system uses OpenCV to find the watch hands and applies geometrical techniques to calculate the time. The second system uses Machine Learning by building a neural network to classify images in Tensorflow using a multi-labelling approach. The results show that in a set environment the geometric-based approach performs better than the Machine Learning model. The geometric system predicted time correctly with an accuracy of 80% whereas the best Machine Learning model only achieves 74%. Experiments show that the accuracy of the neural network model did increase when using data augmentation, however there was no significant improvement when adding synthetic data to our training set.

Advancements in technology, such as smartphones and wearabledevices, can be used for collecting movement data through embeddedsensors. This paper focuses on linking Parkinson’s Disease severitywith data collected from mobile phones in the mPower study. As referencefor symptoms’ severity, we use the answers provided to part 2 ofthe standard MDS-UPDRS scale. As input variables, we use the featurescomputed within mPower from the raw data collected during 4 phonebasedactivities: walking, rest, voice and finger tapping. The features arefiltered in order to remove unreliable datapoints and associated to referencevalues. After pre-processing, 5 Machine Learning algorithms areapplied for predictive analysis. We show that, notwithstanding the noisedue to the data being collected in an uncontrolled manner, the regressedsymptom levels are moderately to strongly correlated with the actualvalues (highest Pearson’s correlation = 0.6). However, the high differencebetween the values also implies that the regressed values can not beconsidered as a substitute for a conventional clinical assessment (lowestmean absolute error = 5.4).

In this paper, we consider the variant of the art gallery problem where the input polygon is a staircase polygon. Previous works on this problem gave a 2-approximation for point guarding a staircase polygon (where guards can be placed anywhere in the interior of the polygon and we wish to guard the entire polygon). It is still unknown whether this point guarding variant is NP-hard. In this paper we consider the vertex guarding problem, where guards are only allowed to be placed at the vertices of the polygon, and we wish to guard only the vertices of the polygon. We show that this problem is NP-hard, and we give a polynomial-time 2-approximation algorithm. 

Human creativity has been often aided and supported by artificial tools, spanning traditional tools such as ideation cards, pens, and paper, to computed and software. Tools for creativity are increasingly using artificial intelligence to not only support the creative process, but also to act upon the creation with a higher level of agency. This paper focuses on writing fiction as a creative activity and explores human-AI co-writing through a research product, which employs a natural language processing model, the Generative Pre-trained Transformer 3 (GPT-3), to assist the co-authoring of narrative fiction. We report on two progressive – not comparative – autoethnographic studies to attain our own creative practices in light of our engagement with the research product: (1) a co-writing activity initiated by basic textual prompts using basic elements of narrative and (2) a co-writing activity initiated by more advanced textual prompts using elements of narrative, including dialects and metaphors undertaken by one of the authors of this paper who has doctoral training in literature. In both studies, we quickly came up against the limitations of the system; then, we repositioned our goals and practices to maximize our chances of success. As a result, we discovered not only limitations but also hidden capabilities, which not only altered our creative practices and outcomes, but which began to change the ways we were relating to the AI as collaborator.  

Recommender Systems (RS) are used in many different applications such as ecommerce and for media streaming, including music. Recommenders not only help users discover new music, but they also help to create assemblages of songs into playlists. Intentionally or otherwise, playlists often manifest themes, that is, universal ideas that are expressed in particular songs, lyrics, or passages. In this paper we were interested to explore the capabilities of AI to introduce themes through generated playlists, them-selves seeded by the theme of plants. Taking a self-reflexive and user experience approach, we collaborated with AI to create four Plant Music playlists to subject ourselves to what came to refer to as a posthuman user experience.  

MnemoRoom4U is an AR (Augmented Reality) tool that uses a memory-palace strategy for foreign-language training. A memory palace helps information recall with the aid of object association in visualizations of familiar spatial surroundings. In MnemoRoom4U, paper or digital flashcards are re-placed with virtual notes containing L1 words and their L2 translations that are placed on top of real physical objects inside a familiar environment, such as one’s room, home, office space, etc. The AR-supported notes aid associative memory by establishing a relationship between the physical objects in the user’s mind and the virtual lexis to be retained in L2. Learners first set up a path through their familiar environment, attaching virtual sticky notes—each containing a target word to be memorized together with its corresponding source-language translation—to real-life objects (e.g. furniture in their homes or offices). They then take the same path again, reviewing all the words, and finally carry out a retention test. MnemoRoom4U is a technological artifact designed for specific didactic purposes in the Unity game engine with the ARCore augmented-reality plug-in for Android. This work takes a Design-Science approach with phenomenological, exploratory underpinnings tracking back to the efficiency of spatial mnemonics previously reported quantitatively and combines it with AR technology to effect L2 vocabulary recall.

Games are complex, multi-faceted systems that share common elements and underlying narratives, such as the conflict between a hero and a big bad enemy or pursuing a goal that requires overcoming challenges. However, identifying and describing these elements together is non-trivial as they might differ in certain properties and how players might encounter the narratives. Likewise, generating narratives also pose difficulties when encoding, interpreting, and evaluating them. To address this, we present TropeTwist, a trope-based system that can describe narrative structures in games in a more abstract and generic level, allowing the definition of games’ narrative structures and their generation using interconnected tropes, called narrative graphs. To demonstrate the system, we represent the narrative structure of three different games. We use MAP-Elites to generate and evaluate novel quality-diverse narrative graphs encoded as graph grammars, using these three hand-made narrative structures as targets. Both hand-made and generated narrative graphs are evaluated based on their coherence and interestingness, which are improved through evolution.  

Narratives are a predominant part of games, and their design poses challenges when identifying, encoding, interpreting, evaluating, and generating them. One way to address this would be to approach narrative design in a more abstract layer, such as narrative structures. This paper presents Story Designer, a mixed-initiative co-creative narrative structure tool built on top of the Evolutionary Dungeon Designer (EDD) that uses tropes, narrative conventions found across many media types, to design these structures. Story Designer uses tropes as building blocks for narrative designers to compose complete narrative structures by interconnecting them in graph structures called narrative graphs. Our mixed-initiative approach lets designers manually create their narrative graphs and feeds an underlying evolutionary algorithm with those, creating quality-diverse suggestions using MAP-Elites. Suggestions are visually represented for designers to compare and evaluate and can then be incorporated into the design for further manual editions. At the same time, we use the levels designed within EDD as constraints for the narrative structure, intertwining both level design and narrative. We evaluate the impact of these constraints and the system’s adaptability and expressiveness, resulting in a potential tool to create narrative structures combining level design aspects with narrative.  

Supporting customers after they acquire the product is essential for companies producing and selling software-intensive embedded systems products. Generally, customer support is the first interaction point between the product users and the product vendor. Customer support is often engaged with answering customers' questions, troubleshooting, fault identification, and fixing product faults. While continuous deployment advocates for closer cooperation between the ones operating the software and the ones developing it, the means of such collaboration in general and the role of customer support, in particular, has not been addressed in the context of software-intensive embedded systems. Therefore, to better understand the impact that continuous deployment has on customer support and the role customer support should play in this context, we conducted a case study at a multinational company developing and selling telecommunications networks infrastructure. We focused on the 4th and 5th Generation (4G and 5G) Radio Access Networks (RAN) products, which can be considered a high volume product as they cover more than 80% of the world's population. Our study reveals that customer support needs to transition from a transaction-based and passive function triggered by customer support requests, to take an active role characterized by being proactive and preemptive to cope with the shorter operational time of a software version introduced by continuous deployment. In addition, customer support plays an essential role in making the feedback actionable by aggregating and consolidating feedback data to the R&D organization.

AI in the Engineering, Procurement and Construction (EPC) industry has not yet a proven track record in large-scale projects. Since AI solutions for industrial applications became available only recently, deployment experience and lessons learned are still to be built up. Several research papers exist describing the potential of AI, and many surveys and white papers have been published indicating the challenges of AI deployment in the EPC industry. However, there is a recognizable shortage of in-depth studies of deployment experience in academic literature, particularly those focusing on the experiences of EPC companies involved in large-scale project execution with high safety standards, such as the petrochemical or energy sector. The novelty of this research is that we explore in detail the challenges and obstacles faced in developing and deploying AI in a large-scale project in the EPC industry based on real-life use cases performed in an EPC company. Those identified challenges are not linked to specific technology or a company's know-how and, therefore, are universal. The findings in this paper aim to provide feedback to academia to reduce the gap between research and practice experience. They also help reveal the hidden stones when implementing AI solutions in the industry.

Internet of Things (IoT) systems are complex systems that can manage mission-critical, costly operations or the collection, storage, and processing of sensitive data. Therefore, security represents a primary concern that should be considered when engineering IoT systems. Additionally, several challenges need to be addressed, including the following ones. IoT systems' environments are dynamic and uncertain. For instance, IoT devices can be mobile or might run out of batteries, so they can become suddenly unavailable. To cope with such environments, IoT systems can be engineered as goal-driven and self-adaptive systems. A goal-driven IoT system is composed of a dynamic set of IoT devices and services that temporarily connect and cooperate to achieve a specific goal. Several approaches have been proposed to engineer goal-driven and self-adaptive IoT systems. However, none of the existing approaches enable goal-driven IoT systems to automatically detect security threats and autonomously adapt to mitigate them. Toward bridging these gaps, this paper proposes a distributed architectural Approach for engineering goal-driven IoT Systems that can autonomously SElf-adapt to secuRity Threats in their environments (ASSERT). ASSERT exploits techniques and adopts notions, such as agents, federated learning, feedback loops, and blockchain, for maintaining the systems' security and enhancing the trustworthiness of the adaptations they perform. The results of the experiments that we conducted to validate the approach's feasibility show that it performs and scales well when detecting security threats, performing autonomous security adaptations to mitigate the threats and enabling systems' constituents to learn about security threats in their environments collaboratively.

Sparrow Search Algorithm (SSA) is a kind of novel swarm intelligence algorithm, which has been applied in-to various domains because of its unique characteristics, such as strong global search capability, few adjustable parameters, and a clear structure. However, the SSA still has some inherent weaknesses that hinder its further development, such as poor population diversity, weak local searchability, and falling into local optima easily. This manuscript proposes an improved chaos sparrow search optimization algorithm (ICSSOA) to overcome the mentioned shortcomings of the standard SSA. Firstly, the Cubic chaos mapping is introduced to increase the population diversity in the initialization stage. Then, an adaptive weight is employed to automatically adjust the search step for balancing the global search performance and the local search capability in different phases. Finally, a hybrid strategy of Levy flight and reverse learning is presented to perturb the position of individuals in the population according to the random strategy, and a greedy strategy is utilized to select individuals with higher fitness values to decrease the possibility of falling into the local optimum. The experiments are divided into two modules. The former investigates the performance of the proposed approach through 20 benchmark functions optimization using the ICSSOA, standard SSA, and other four SSA variants. In the latter experiment, the selected 20 functions are also optimized by the ICSSOA and other classic swarm intelligence algorithms, namely ACO, PSO, GWO, and WOA. Experimental results and corresponding statistical analysis revealed that only one function optimization test using the ICSSOA was slightly lower than the CSSOA and the WOA among the 20-function optimization. In most cases, the values for both accuracy and convergence speed are higher than other algorithms. The results also indicate that the ICSSOA has an outstanding ability to jump out of the local optimum.

Game design is a hard and multi-faceted task that intertwines different gameplay mechanics, audio, level, graphic, and narrative facets. Games' facets are developed in conjunction with others with a common goal that makes games coherent and interesting. These combinations result in plenty of games in diverse genres, which usually require a collaboration of a diverse group of designers. Collaborators can take different roles and support each other with their strengths resulting in games with unique characteristics. The multi-faceted nature of games and their collaborative properties and requirements make it an exciting task to use Artificial Intelligence (AI). The generation of these facets together requires a holistic approach, which is one of the most challenging tasks within computational creativity. Given the collaborative aspect of games, this thesis approaches their generation through Human-AI collaboration, specifically using a mixed-initiative co-creative (MI-CC) paradigm. This paradigm creates an interactive and collaborative scenario that leverages AI and human strengths with an alternating and proactive initiative to approach a task. However, this paradigm introduces several challenges, such as Human and AI goal alignment or competing properties.

In this thesis, game design and the generation of game facets by themselves and intertwined are explored through Human-AI collaboration. The AI takes a colleague's role with the designer, arising multiple dynamics, challenges, and opportunities. The main hypothesis is that AI can be incorporated into systems as a collaborator, enhancing design tools, fostering human creativity, and reducing workload. The challenges and opportunities that arise from this are explored, discussed, and approached throughout the thesis. As a result, multiple approaches and methods such as quality-diversity algorithms and designer modeling are proposed to generate game facets in tandem with humans, create a better workflow, enhance the interaction, and establish adaptive experiences.

In Mixed-Initiative Co-Creative tools, the human is mostly in control of what will and can be created, delegating the AI to a more suggestive role instead of a colleague in the co-creative process. Allowing more control and agency for the AI might be an interesting path in co-creative scenarios where AI could direct and take more initiative within the co-creative task. However, the relationship between AI and human designers in creative processes is delicate, as adjusting the initiative or agency of the AI can negatively affect the user experience. In this paper, different degrees of agency for the AI are explored within the Evolutionary Dungeon Designer (EDD) to further understand MI-CC tools. A user study was performed using EDD with three varying degrees of AI agency. The study highlighted elements of frustration that the human designer experiences when using the tool and the behavior in the AI that led to possible strains on the relationship. The paper concludes with the identified issues and possible solutions and suggested further research.

Games are complex, multi-faceted systems that share common elements and underlying narratives, such as the conflict between a hero and a big bad enemy or pursuing a goal that requires overcoming challenges. However, identifying and describing these elements together is non-trivial as they might differ in certain properties and how players might encounter the narratives. Likewise, generating narratives also pose difficulties when encoding, interpreting, and evaluating them. To address this, we present TropeTwist, a trope-based system that can describe narrative structures in games in a more abstract and generic level, allowing the definition of games' narrative structures and their generation using interconnected tropes, called narrative graphs. To demonstrate the system, we represent the narrative structure of three different games. We use MAP-Elites to generate and evaluate novel quality-diverse narrative graphs encoded as graph grammars, using these three hand-made narrative structures as targets. Both hand-made and generated narrative graphs are evaluated based on their coherence and interestingness, which are improved through evolution.

With the Internet of Things paradigm, the data generated by the rapidly increasing number of connected devices lead to new possibilities, such as using machine learning for activity recognition in smart environments. However, it also introduces several challenges. The sensors of different devices might be mobile and of different types, i.e. there is a need to handle streaming data from a dynamic and heterogeneous set of sensors. In machine learning, the performance is often linked to the availability and quality of annotated data. Annotating data is in general costly, but it can be even more challenging if there is not any, or a very small amount of, annotated data to train the model on at the start of learning. To handle these issues, we implement interactive and adaptive systems. By including human-in-the-loop, which we refer to as interactive machine learning, the input from users can be utilized to build the model. The type of input used in interactive machine learning is typically annotations of the data, i.e. correctly labelled data points. Generally, it is assumed that the user always provides correct labels in accordance with the chosen interactive learning strategy. In many real-world applications these assumptions are not realistic however, as users might provide incorrect labels or not provide labels at all in line with the chosen strategy.

In this thesis we explore which interactive learning strategy types are possible in the given scenario and how they affect performance, as well as the effect of machine learning algorithms on the performance. We also study how a user who is not always reliable, i.e. who does not always provide a correct label when expected to, can affect performance. We propose a taxonomy of interactive online machine learning strategies and test how the different strategies affect performance through experiments on multiple datasets. Simulated experiments are compared to experiments with human participants, to verify the results. The findings show that the overall best performing interactive learning strategy is one where the user provides labels when current estimations are incorrect, but that the best performing machine learning algorithm depends on the problem scenario. The experiments also show that a decreased reliability of the user leads to decreased performance, especially when there is a limited amount of labelled data. The robustness of the machine learning algorithms differs, where e.g. Naïve Bayes classifier is better at handling a lower reliability of the user. We also present a systematic literature review on machine teaching, a subfield of interactive machine learning where the human is proactive in the interaction. The study shows that the area of machine teaching is rapidly evolving with an increased number of publications in recent years. However, as it is still maturing, there exists several open challenges that would benefit from further exploration, e.g. how human factors can affect performance.

The use of artificial intelligence and machine learning techniques across all disciplines has exploded in the past few years, with the ever-growing size of data and the changing needs of higher education, such as digital education. Similarly, online educational information systems have a huge amount of data related to students in digital education. This educational data can be used with artificial intelligence and machine learning techniques to improve digital education. This study makes two main contributions. First, the study follows a repeatable and objective process of exploring the literature. Second, the study outlines and explains the literature's themes related to the use of AI-based algorithms in digital education. The study findings present six themes related to the use of machines in digital education. The synthesized evidence in this study suggests that machine learning and deep learning algorithms are used in several themes of digital learning. These themes include using intelligent tutors, dropout predictions, performance predictions, adaptive and predictive learning and learning styles, analytics and group-based learning, and automation. artificial neural network and support vector machine algorithms appear to be utilized among all the identified themes, followed by random forest, decision tree, naive Bayes, and logistic regression algorithms.

With the increased availability of new and better computer processing units (CPUs) as well as graphical processing units (GPUs), the interest in statistical learning and deep learning algorithms for classification tasks has grown exponentially. These classification algorithms often require the presence of fully labeled instances during the training period for maximum classification accuracy. However, in industrial applications, data is commonly not fully labeled, which both reduces the prediction accuracy of the learning algorithms as well as increases the project cost to label the missing instances. The purpose of this paper is to survey the current state-of-the-art literature on machine learning algorithms that are used for assisted or automatic labeling and to understand where these are used. We performed a systematic mapping study and identified 52 primary studies relevant to our research. This paper provides three main contributions. First, we identify the existing machine learning algorithms for labeling and we present a taxonomy of these algorithms. Second, we identify the datasets that are used to evaluate the algorithms and we provide a mapping of the datasets based on the type of data and the application area. Third, we provide a process to support people in industry to optimally label their dataset. The results presented in this paper can be used by both researchers and practitioners aiming to improve the missing labels with the aid of machine algorithms or to select appropriate datasets to compare new state-of-the art algorithms in their respective application area.

We propose modeling designer style in mixed-initiative game content creation tools as archetypical design traces. These design traces are formulated as transitions between design styles; these design styles are in turn found through clustering all intermediate designs along the way to making a complete design. This method is implemented in the Evolutionary Dungeon Designer, a research platform for mixed-initiative systems to create adventure and dungeon crawler games. We present results both in the form of design styles for rooms, which can be analyzed to better understand the kind of rooms designed by users, and in the form of archetypical sequences between these rooms, i.e., Designer Personas.