Search

Epilepsy affects approximately 50 million people worldwide. Despite its prevalence, the recurrence of seizures can be mitigated only 70% of the time through medication. Furthermore, surgery success rates range from 30% - 70% because of our limited understanding of how a seizure starts. However, one leading hypothesis suggests that a seizure starts because of a critical transition due to an instability. Unfortunately, we lack a meaningful way to quantify this notion that would allow physicians to not only better predict seizures but also to mitigate them. Hence, in this paper, we develop a method to not only characterize the instability of seizures but also to leverage these conditions to stabilize the system underlying these seizures. Remarkably, evidence suggests that such critical transitions are associated with long-term memory dynamics, which can be captured by considering linear fractional-order systems. Subsequently, we provide for the first time tractable necessary and sufficient conditions for the global asymptotic stability of discrete-time linear fractional-order systems. Next, we propose a method to obtain a stabilizing control strategy for these systems using linear matrix inequalities. Finally, we apply our methodology to a real-world epileptic patient dataset to provide insight into mitigating epilepsy and designing future cyber-neural systems.

The impulsive Goodwin's oscillator (IGO) is a hybrid model composed of a third-order continuous linear part and a pulse-modulated feedback. This paper introduces a design problem of the IGO to admit a desired periodic solution. The dynamics of the continuous states represent the plant to be controlled, whereas the parameters of the impulsive feedback constitute design degrees of freedom. The design objective is to select the free parameters so that the IGO exhibits a stable 1-cycle with desired characteristics. The impulse-to-impulse map of the oscillator is demonstrated to always possess a positive fixed point that corresponds to the desired periodic solution; the closed-form expressions to evaluate this fixed point are provided. Necessary and sufficient conditions for orbital stability of the 1-cycle are presented in terms of the oscillator parameters and exhibit similarity to the problem of static output control. An IGO design procedure is proposed and validated by simulation. The nonlinear dynamics of the designed IGO are reviewed by means of bifurcation analysis. Applications of the design procedure to dosing problems in chemical industry and biomedicine are envisioned.

There has recently been an increased interest in reinforcement learning for nonlinear control problems. However standard reinforcement learning algorithms can often struggle even on seemingly simple set-point control problems. This paper argues that three ideas can improve reinforcement learning methods even for highly nonlinear set-point control problems: 1) Make use of a prior feedback controller to aid amplitude exploration. 2) Use integrated errors. 3) Train on model ensembles. Together these ideas lead to more efficient training, and a trained set-point controller that is more robust to modelling errors and thus can be directly deployed to real-world nonlinear systems. The claim is supported by experiments with a real-world nonlinear cascaded tank process and a simulated strongly nonlinear pH-control system.

This paper considers the task of learning how to make a prognosis of a patient based on his/her micro-array expression levels. The method is an application of the aggregation method as recently proposed in the literature on theoretical machine learning, and excels in its computational convenience and capability to deal with high-dimensional data. This paper gives a formal analysis of the method, yielding rates of convergence similar to what traditional techniques obtain, while it is shown to cope well with an exponentially large set of features. Those results are supported by numerical simulations on a range of publicly available survival-micro-array data sets. It is empirically found that the proposed technique combined with a recently proposed pre-processing technique gives excellent performances. All used software files and data sets are available on the authors' website http://user.it.uu.se/similar to liuya610/index.html

Recent developments within deep learning are relevant for nonlinear system identification problems. In this paper, we establish connections between the deep learning and the system identification communities. It has recently been shown that convolutional architectures are at least as capable as recurrent architectures when it comes to sequence modeling tasks. Inspired by these results we explore the explicit relationships between the recently proposed temporal convolutional network (TCN) and two classic system identification model structures; Volterra series and block-oriented models. We end the paper with an experimental study where we provide results on two real-world problems, the well-known Silverbox dataset and a newer dataset originating from ground vibration experiments on an F-16 fighter aircraft.

In this paper, we perform structural analyses of a parsimonious watermarking policy, which minimizes the average detection delay (ADD) to detect data deception attacks on networked control systems (NCS) for a fixed upper bound on the false alarm rate (FAR). The addition of physical watermarking to the control input of a NCS increases the probability of attack detections with an increase in the control cost. Therefore, we formulate the problem of data deception attack detection for NCS with the facility to add physical watermarking as a stochastic optimal control problem. Then we solve the problem by applying dynamic programming value iterations and find a parsimonious watermarking policy that decides to add watermarking and detects attacks based on the estimated posterior probability of attack. We analyze the optimal policy structure and find that it can be a one, two or three threshold policy depending on a few parameter values. Simulation studies show that the optimal policy for a practical range of parameter values is a two-threshold policy on the posterior probability of attack. Derivation of a threshold-based policy from the structural analysis of the value iteration method reduces the computational complexity during the runtime implementation and offers better structural insights. Furthermore, such an analysis provides a guideline for selecting the parameter values to meet the design requirements.

Antenatal depression impacts 7-20% of women globally, and can have serious consequences for both the mother and the infant. Preventative interventions are effective, but are cost-efficient only among those at high risk. As such, being able to predict and identify those at risk is invaluable for reducing the burden of care and adverse consequences, as well as improving treatment outcomes. While several approaches have been proposed in the literature for the automatic prediction of depressive states, there is a scarcity of research on automatic prediction of perinatal depression. Moreover, while there exist some works on the automatic prediction of postpartum depression using data collected in clinical settings and applied the model to a smartphone application, to the best of our knowledge, no previous work has investigated the automatic prediction of late antenatal depression using data collected via a smartphone app in the first and second trimesters of pregnancy. This study utilizes data measuring various aspects of self-reported psychological, physiological and behavioral information, collected from 915 women in the first and second trimesters of pregnancy using a smartphone app designed for perinatal depression. By applying machine learning algorithms on these data, this paper explores the possibility of automatic early detection of antenatal depression (i.e., during week 36 to week 42 of pregnancy) in everyday life without the administration of healthcare professionals. We compare uni-modal and multi-modal models and identify predictive markers related to antenatal depression. With multi-modal approach the model reaches a BAC of 0.75, and an AUC of 0.82.

Antenatal depression impacts 7-20% of women globally, and can have serious consequences for both the mother and the infant. Preventative interventions are effective, but are cost-efficient only among those at high risk. As such, being able to predict and identify those at risk is invaluable for reducing the burden of care and adverse consequences, as well as improving treatment outcomes. While several approaches have been proposed in the literature for the automatic prediction of depressive states, there is a scarcity of research on automatic prediction of perinatal depression. Moreover, while there exist some works on the automatic prediction of postpartum depression using data collected in clinical settings and applied the model to a smartphone application, to the best of our knowledge, no previous work has investigated the automatic prediction of late antenatal depression using data collected via a smartphone app in the first and second trimesters of pregnancy. This study utilizes data measuring various aspects of self-reported psychological, physiological and behavioral information, collected from 915 women in the first and second trimesters of pregnancy using a smartphone app designed for perinatal depression. By applying machine learning algorithms on these data, this paper explores the possibility of automatic early detection of antenatal depression (i.e., during week 36 to week 42 of pregnancy) in everyday life without the administration of healthcare professionals. We compare uni-modal and multi-modal models and identify predictive markers related to antenatal depression. With multi-modal approach the model reaches a BAC of 0.75, and an AUC of 0.82.

This paper introduces a stochastic framework for a recently proposed discrete-time delay estimation method in Laguerre-domain, i.e. with the delay block input and output signals being represented by the corresponding Laguerre series. A novel Laguerre-domain disturbance model allowing the involved signals to be square-summable sequences is devised. The relation to two commonly used time-domain disturbance models is clarified. Furthermore, by forming the input signal in a certain way, the signal shape of an additive output disturbance can be estimated and utilized for noise reduction. It is demonstrated that a significant improvement in the delay estimation error is achieved when the noise sequence is correlated. The noise reduction approach is applicable to other Laguerre-domain problems than pure delay estimation.

This paper introduces a stochastic framework for a recently proposed discrete-time delay estimation method in Laguerre-domain, i.e. with the delay block input and output signals being represented by the corresponding Laguerre series. A novel Laguerre domain disturbance model is devised, which allows the involved signals to be square-summable sequences and is suitable in a number of important applications. The relation to two commonly used time-domain disturbance models is clarified. Furthermore, by forming the input signal in a certain way, the signal shape of an additive output disturbance can be estimated and utilized for noise reduction. It is demonstrated that a significant improvement in the delay estimation error is achieved when the noise sequence is correlated. The noise reduction approach is applicable to other Laguerre-domain problems than pure delay estimation.

Differential-algebraic equations, commonly used to model physical systems, are the basis for many equation-based object-oriented modeling languages. When systems described by such equations are influenced by unknown process disturbances, estimating unknown parameters from experimental data becomes difficult. This is because of problems with the existence of well-defined solutions and the computational tractability of estimators. In this paper, we propose a way to minimize a cost function-whose minimizer is a consistent estimator of the true parameters-using stochastic gradient descent. This approach scales significantly better with the number of unknown parameters than other currently available methods for the same type of problem. The performance of the method is demonstrated through a simulation study with three unknown parameters. The experiments show a significantly reduced variance of the estimator, compared to an output error method neglecting the influence of process disturbances. The proposed approach is also able to reduce the estimation bias of parameters that the output error method particularly struggles with.

This thesis is centred around Gaussian processes and large-scale optimisation, where the main contributions are presented in the included papers. Provided access to linear constraints (e.g. equilibrium conditions), we propose a constructive procedure to design the covariance function in a Gaussian process. The constraints are thereby explicitly incorporated with guaranteed fulfilment. One such construction is successfully applied to strain field reconstruction, where the goal is to describe the interior of a deformed object. Furthermore, we analyse the Gaussian process as a tool for X-ray computed tomography, a field of high importance primarily due to its central role in medical treatments. This provides insightful interpretations of traditional reconstruction algorithms. Large-scale optimisation is considered in two different contexts. First, we consider a stochastic environment, for which we suggest a new method inspired by the quasi-Newton framework. Promising results are demonstrated on real world benchmark problems. Secondly, we suggest an approach to solve an applied deterministic optimisation problem that arises within the design of electrical circuit boards. We reduce the memory requirements through a tailored algorithm, while also benefiting from other parts of the setting to ensure a high computational efficiency. The final paper scrutinises a publication from the early phase of the COVID-19 pandemic, in which the aim was to assess the effectiveness of different governmental interventions. We show that minor modifications in the input data have important impact on the results, and we argue that great caution is necessary when such models are used as a support for decision making.

This thesis is centred around Gaussian processes and large-scale optimisation, where the main contributions are presented in the included papers. Provided access to linear constraints (e.g. equilibrium conditions), we propose a constructive procedure to design the covariance function in a Gaussian process. The constraints are thereby explicitly incorporated with guaranteed fulfilment. One such construction is successfully applied to strain field reconstruction, where the goal is to describe the interior of a deformed object. Furthermore, we analyse the Gaussian process as a tool for X-ray computed tomography, a field of high importance primarily due to its central role in medical treatments. This provides insightful interpretations of traditional reconstruction algorithms. Large-scale optimisation is considered in two different contexts. First, we consider a stochastic environment, for which we suggest a new method inspired by the quasi-Newton framework. Promising results are demonstrated on real world benchmark problems. Secondly, we suggest an approach to solve an applied deterministic optimisation problem that arises within the design of electrical circuit boards. We reduce the memory requirements through a tailored algorithm, while also benefiting from other parts of the setting to ensure a high computational efficiency. The final paper scrutinises a publication from the early phase of the COVID-19 pandemic, in which the aim was to assess the effectiveness of different governmental interventions. We show that minor modifications in the input data have important impact on the results, and we argue that great caution is necessary when such models are used as a support for decision making.

This paper proposes a game-theoretic approach to address the problem of optimal sensor placement against an adversary in uncertain networked control systems. The problem is formulated as a zero-sum game with two players, namely a malicious adversary and a detector. Given a protected performance vertex, we consider a detector, with uncertain system knowledge, that selects another vertex on which to place a sensor and monitors its output with the aim of detecting the presence of the adversary. On the other hand, the adversary, also with uncertain system knowledge, chooses a single vertex and conducts a cyber-attack on its input. The purpose of the adversary is to drive the attack vertex as to maximally disrupt the protected performance vertex while remaining undetected by the detector. As our first contribution, the game payoff of the above-defined zero-sum game is formulated in terms of the Value-at-Risk of the adversary’s impact. However, this game payoff corresponds to an intractable optimization problem. To tackle the problem, we adopt the scenario approach to approximately compute the game payoff. Then, the optimal monitor selection is determined by analyzing the equilibrium of the zero-sum game. The proposed approach is illustrated via a numerical example of a 10-vertex networked control system.

In this paper, we have proposed a technique for Bayesian sequential detection of replay attacks on networked control systems with a constraint on the average number of watermarking (ANW) events used during normal system operations. Such a constraint limits the increase in the control cost due to watermarking. To determine the optimal sequence regarding the addition or otherwise of watermarking signals, first, we formulate an infinite horizon stochastic optimal control problem with a termination state. Then applying the value iteration approach, we find an optional policy that minimizes the average detection delay (ADD) for fixed upper bounds on the false alarm rate (FAR) and ANW. The optimal policy turns out to be a two thresholds policy on the posterior probability of attack. We derive approximate expressions of ADD and FAR as functions of the two derived thresholds and a few other parameters. A simulation study on a single-input single-output system illustrates that the proposed method improves the control cost considerably at the expense of small increases in ADD. We also perform simulation studies to validate the derived theoretical results.

Antenatal depression impacts 7-20% of women globally, and can have serious consequences for both the mother and the infant. Preventative interventions are effective, but are cost-efficient only among those at high risk. As such, being able to predict and identify those at risk is invaluable for reducing the burden of care and adverse consequences, as well as improving treatment outcomes. While several approaches have been proposed in the literature for the automatic prediction of depressive states, there is a scarcity of research on automatic prediction of perinatal depression. Moreover, while there exist some works on the automatic prediction of postpartum depression using data collected in clinical settings and applied the model to a smartphone application, to the best of our knowledge, no previous work has investigated the automatic prediction of late antenatal depression using data collected via a smartphone app in the first and second trimesters of pregnancy. This study utilizes data measuring various aspects of self-reported psychological, physiological and behavioral information, collected from 915 women in the first and second trimesters of pregnancy using a smartphone app designed for perinatal depression. By applying machine learning algorithms on these data, this paper explores the possibility of automatic early detection of antenatal depression (i.e., during week 36 to week 42 of pregnancy) in everyday life without the administration of healthcare professionals. We compare uni-modal and multi-modal models and identify predictive markers related to antenatal depression. With multi-modal approach the model reaches a BAC of 0.75, and an AUC of 0.82.

When recorded signals are corrupted by noise on both input and output sides, standard identification methods give biased parameter estimates, due to the presence of input noise. This paper discusses in what situations such a bias is large and, consequently, when errors-in-variables identification methods should preferably be used.

High sensor accuracy is reliant on timely calibrations where any drift is corrected for. Common validation practice involves cross-checking the sensor with a grab sample. In this paper we simulated and evaluated the effectiveness of this common approach. The results showed that more than one grab sample was needed to detect a bias of about ten percent. The random uncertainties in the grab sample could be reduced be replicate samples, which thereby improved the bias detection. The findings indicate that careful, at the cost of less frequent validations, is recommended, as compared to the opposite and current practice.

The paper considers the problem of multi-objective decision support when outcomes are uncertain. We extend the concept of Pareto-efficient decisions to take into account the uncertainty of decision outcomes across varying contexts. This enables quantifying trade-offs between decisions in terms of tail outcomes that are relevant in safety-critical applications. We propose a method for learning efficient decisions with statistical confidence, building on results from the conformal prediction literature. The method adapts to weak or nonexistent context covariate overlap and its statistical guarantees are evaluated using both synthetic and real data.

In this paper, we consider the optimal controller design problem against data injection attacks on actuators for an uncertain control system. We consider attacks that aim at maximizing the attack impact while remaining stealthy in the finite horizon. To this end, we use the Conditional Value-at-Risk to characterize the risk associated with the impact of attacks. The worst-case attack impact is characterized using the recently proposed output-to-output l(2)-gain (OOG). We formulate the design problem and observe that it is non-convex and hard to solve. Using the framework of scenariobased optimization and a convex proxy for the OOG, we propose a convex optimization problem that approximately solves the design problem with probabilistic certificates. Finally, we illustrate the results through a numerical example.

Antenatal depression impacts 7-20% of women globally, and can have serious consequences for both the mother and the infant. Preventative interventions are effective, but are cost-efficient only among those at high risk. As such, being able to predict and identify those at risk is invaluable for reducing the burden of care and adverse consequences, as well as improving treatment outcomes. While several approaches have been proposed in the literature for the automatic prediction of depressive states, there is a scarcity of research on automatic prediction of perinatal depression. Moreover, while there exist some works on the automatic prediction of postpartum depression using data collected in clinical settings and applied the model to a smartphone application, to the best of our knowledge, no previous work has investigated the automatic prediction of late antenatal depression using data collected via a smartphone app in the first and second trimesters of pregnancy. This study utilizes data measuring various aspects of self-reported psychological, physiological and behavioral information, collected from 915 women in the first and second trimesters of pregnancy using a smartphone app designed for perinatal depression. By applying machine learning algorithms on these data, this paper explores the possibility of automatic early detection of antenatal depression (i.e., during week 36 to week 42 of pregnancy) in everyday life without the administration of healthcare professionals. We compare uni-modal and multi-modal models and identify predictive markers related to antenatal depression. With multi-modal approach the model reaches a BAC of 0.75, and an AUC of 0.82.

Antenatal depression impacts 7-20% of women globally, and can have serious consequences for both the mother and the infant. Preventative interventions are effective, but are cost-efficient only among those at high risk. As such, being able to predict and identify those at risk is invaluable for reducing the burden of care and adverse consequences, as well as improving treatment outcomes. While several approaches have been proposed in the literature for the automatic prediction of depressive states, there is a scarcity of research on automatic prediction of perinatal depression. Moreover, while there exist some works on the automatic prediction of postpartum depression using data collected in clinical settings and applied the model to a smartphone application, to the best of our knowledge, no previous work has investigated the automatic prediction of late antenatal depression using data collected via a smartphone app in the first and second trimesters of pregnancy. This study utilizes data measuring various aspects of self-reported psychological, physiological and behavioral information, collected from 915 women in the first and second trimesters of pregnancy using a smartphone app designed for perinatal depression. By applying machine learning algorithms on these data, this paper explores the possibility of automatic early detection of antenatal depression (i.e., during week 36 to week 42 of pregnancy) in everyday life without the administration of healthcare professionals. We compare uni-modal and multi-modal models and identify predictive markers related to antenatal depression. With multi-modal approach the model reaches a BAC of 0.75, and an AUC of 0.82.

This thesis is centred around Gaussian processes and large-scale optimisation, where the main contributions are presented in the included papers. Provided access to linear constraints (e.g. equilibrium conditions), we propose a constructive procedure to design the covariance function in a Gaussian process. The constraints are thereby explicitly incorporated with guaranteed fulfilment. One such construction is successfully applied to strain field reconstruction, where the goal is to describe the interior of a deformed object. Furthermore, we analyse the Gaussian process as a tool for X-ray computed tomography, a field of high importance primarily due to its central role in medical treatments. This provides insightful interpretations of traditional reconstruction algorithms. Large-scale optimisation is considered in two different contexts. First, we consider a stochastic environment, for which we suggest a new method inspired by the quasi-Newton framework. Promising results are demonstrated on real world benchmark problems. Secondly, we suggest an approach to solve an applied deterministic optimisation problem that arises within the design of electrical circuit boards. We reduce the memory requirements through a tailored algorithm, while also benefiting from other parts of the setting to ensure a high computational efficiency. The final paper scrutinises a publication from the early phase of the COVID-19 pandemic, in which the aim was to assess the effectiveness of different governmental interventions. We show that minor modifications in the input data have important impact on the results, and we argue that great caution is necessary when such models are used as a support for decision making.

This paper presents a literature review and interview study exploring the opportunities and hurdles when implementing Artificial Intelligence (AI) in agricultural businesses. Three sectors in agriculture are scrutinized: arable farming, milk production and beef production. As a foundation for the project, a literature review revises former research on smart farming. Thereafter, an interview study with 21 respondents both explores the susceptibility and maturity of smart farming technologies and provides technical depth to three chosen applications of AI in agriculture. Although the focus of the study is on the Swedish context, the findings can be generalized to the European agricultural sector and even world-wide. Findings of the study include a diverse set of aspects that both enable and obstruct the transition. Main identified opportunities are the importance smart farming has on the strategic agendas of several industry stakeholders, the general trend towards software technology as a service through shared machinery, the vast amount of existing data, and the large interest from farmers towards new technology. Contrasting, the study identifies main hurdles as technical and legislative challenges to data ownership, potential cybersecurity threats, the need for a well-articulated business case, and the sometimes lacking technical knowledge within the sector. The paper concludes that the macro trend points towards a smart farming transition but that the speed of the transformation will depend on the resolutions for the identified obstacles.

This thesis is centred around Gaussian processes and large-scale optimisation, where the main contributions are presented in the included papers. Provided access to linear constraints (e.g. equilibrium conditions), we propose a constructive procedure to design the covariance function in a Gaussian process. The constraints are thereby explicitly incorporated with guaranteed fulfilment. One such construction is successfully applied to strain field reconstruction, where the goal is to describe the interior of a deformed object. Furthermore, we analyse the Gaussian process as a tool for X-ray computed tomography, a field of high importance primarily due to its central role in medical treatments. This provides insightful interpretations of traditional reconstruction algorithms. Large-scale optimisation is considered in two different contexts. First, we consider a stochastic environment, for which we suggest a new method inspired by the quasi-Newton framework. Promising results are demonstrated on real world benchmark problems. Secondly, we suggest an approach to solve an applied deterministic optimisation problem that arises within the design of electrical circuit boards. We reduce the memory requirements through a tailored algorithm, while also benefiting from other parts of the setting to ensure a high computational efficiency. The final paper scrutinises a publication from the early phase of the COVID-19 pandemic, in which the aim was to assess the effectiveness of different governmental interventions. We show that minor modifications in the input data have important impact on the results, and we argue that great caution is necessary when such models are used as a support for decision making.

This paper presents a literature review and interview study exploring the opportunities and hurdles when implementing Artificial Intelligence (AI) in agricultural businesses. Three sectors in agriculture are scrutinized: arable farming, milk production and beef production. As a foundation for the project, a literature review revises former research on smart farming. Thereafter, an interview study with 21 respondents both explores the susceptibility and maturity of smart farming technologies and provides technical depth to three chosen applications of AI in agriculture. Although the focus of the study is on the Swedish context, the findings can be generalized to the European agricultural sector and even world-wide. Findings of the study include a diverse set of aspects that both enable and obstruct the transition. Main identified opportunities are the importance smart farming has on the strategic agendas of several industry stakeholders, the general trend towards software technology as a service through shared machinery, the vast amount of existing data, and the large interest from farmers towards new technology. Contrasting, the study identifies main hurdles as technical and legislative challenges to data ownership, potential cybersecurity threats, the need for a well-articulated business case, and the sometimes lacking technical knowledge within the sector. The paper concludes that the macro trend points towards a smart farming transition but that the speed of the transformation will depend on the resolutions for the identified obstacles.

This paper presents a literature review and interview study exploring the opportunities and hurdles when implementing Artificial Intelligence (AI) in agricultural businesses. Three sectors in agriculture are scrutinized: arable farming, milk production and beef production. As a foundation for the project, a literature review revises former research on smart farming. Thereafter, an interview study with 21 respondents both explores the susceptibility and maturity of smart farming technologies and provides technical depth to three chosen applications of AI in agriculture. Although the focus of the study is on the Swedish context, the findings can be generalized to the European agricultural sector and even world-wide. Findings of the study include a diverse set of aspects that both enable and obstruct the transition. Main identified opportunities are the importance smart farming has on the strategic agendas of several industry stakeholders, the general trend towards software technology as a service through shared machinery, the vast amount of existing data, and the large interest from farmers towards new technology. Contrasting, the study identifies main hurdles as technical and legislative challenges to data ownership, potential cybersecurity threats, the need for a well-articulated business case, and the sometimes lacking technical knowledge within the sector. The paper concludes that the macro trend points towards a smart farming transition but that the speed of the transformation will depend on the resolutions for the identified obstacles.

This paper presents a literature review and interview study exploring the opportunities and hurdles when implementing Artificial Intelligence (AI) in agricultural businesses. Three sectors in agriculture are scrutinized: arable farming, milk production and beef production. As a foundation for the project, a literature review revises former research on smart farming. Thereafter, an interview study with 21 respondents both explores the susceptibility and maturity of smart farming technologies and provides technical depth to three chosen applications of AI in agriculture. Although the focus of the study is on the Swedish context, the findings can be generalized to the European agricultural sector and even world-wide. Findings of the study include a diverse set of aspects that both enable and obstruct the transition. Main identified opportunities are the importance smart farming has on the strategic agendas of several industry stakeholders, the general trend towards software technology as a service through shared machinery, the vast amount of existing data, and the large interest from farmers towards new technology. Contrasting, the study identifies main hurdles as technical and legislative challenges to data ownership, potential cybersecurity threats, the need for a well-articulated business case, and the sometimes lacking technical knowledge within the sector. The paper concludes that the macro trend points towards a smart farming transition but that the speed of the transformation will depend on the resolutions for the identified obstacles.

This thesis is centred around Gaussian processes and large-scale optimisation, where the main contributions are presented in the included papers. Provided access to linear constraints (e.g. equilibrium conditions), we propose a constructive procedure to design the covariance function in a Gaussian process. The constraints are thereby explicitly incorporated with guaranteed fulfilment. One such construction is successfully applied to strain field reconstruction, where the goal is to describe the interior of a deformed object. Furthermore, we analyse the Gaussian process as a tool for X-ray computed tomography, a field of high importance primarily due to its central role in medical treatments. This provides insightful interpretations of traditional reconstruction algorithms. Large-scale optimisation is considered in two different contexts. First, we consider a stochastic environment, for which we suggest a new method inspired by the quasi-Newton framework. Promising results are demonstrated on real world benchmark problems. Secondly, we suggest an approach to solve an applied deterministic optimisation problem that arises within the design of electrical circuit boards. We reduce the memory requirements through a tailored algorithm, while also benefiting from other parts of the setting to ensure a high computational efficiency. The final paper scrutinises a publication from the early phase of the COVID-19 pandemic, in which the aim was to assess the effectiveness of different governmental interventions. We show that minor modifications in the input data have important impact on the results, and we argue that great caution is necessary when such models are used as a support for decision making.

This paper presents a literature review and interview study exploring the opportunities and hurdles when implementing Artificial Intelligence (AI) in agricultural businesses. Three sectors in agriculture are scrutinized: arable farming, milk production and beef production. As a foundation for the project, a literature review revises former research on smart farming. Thereafter, an interview study with 21 respondents both explores the susceptibility and maturity of smart farming technologies and provides technical depth to three chosen applications of AI in agriculture. Although the focus of the study is on the Swedish context, the findings can be generalized to the European agricultural sector and even world-wide. Findings of the study include a diverse set of aspects that both enable and obstruct the transition. Main identified opportunities are the importance smart farming has on the strategic agendas of several industry stakeholders, the general trend towards software technology as a service through shared machinery, the vast amount of existing data, and the large interest from farmers towards new technology. Contrasting, the study identifies main hurdles as technical and legislative challenges to data ownership, potential cybersecurity threats, the need for a well-articulated business case, and the sometimes lacking technical knowledge within the sector. The paper concludes that the macro trend points towards a smart farming transition but that the speed of the transformation will depend on the resolutions for the identified obstacles.

We consider the problem of sinusoidal parameter estimation using signed observations obtained via one-bit sampling with time-varying thresholds. In a previous paper, a relaxation based algorithm, referred to as IbRELAX, has been proposed to iteratively maximize the likelihood function. However, IbRELAX can only he used in applications involving a small number of sinusoids due to the time-consuming exhaustive search procedure needed in each iteration. In this paper, we present a majorizalion-minimization (MM) based IbRELAX algorithm, referred to as IbMMRELAX, to enhance the computational efficiency of IbRELAX. Using the MM technique, IbMMRELAX maximizes the likelihood function iteratively using simple FFT operations to reduce the computational cost of IbRELAX while maintaining its excellent estimation accuracy. Numerical examples are presented to demonstrate the effectiveness of the proposed method.

This paper studies a condition for global parametric convergence of a recursive identification algorithm with an output quantizer. The problem is addressed by an analytic calculation of the associated ODE for a low order test model. The ODE is then analysed numerically to investigate the possibilities for proving parametric convergence. The main contribution of the paper shows that a monotone quantizer is not a necessary condition for global parametric convergence to the true parameter vector w.p.1.

The impulsive Goodwin's oscillator (IGO) is a hybrid model that captures complex dynamics arising in continuous systems controlled by pulse-modulated (event-based) feedback. Being conceived to describe pulsatile endocrine regulation, it has also found applications in e.g. pharmacokinetics. The original version of the IGO assumes the continuous part of the model to be a chain of first-order blocks. This paper explores the nonlinear phenomena arising due to the introduction of a local continuous feedback as suggested by the endocrine applications. The effects caused by a nonlinear feedback law parameterized by a Hill function are compared to those arising due to a simpler and previously treated case of affine feedback law. The hybrid dynamics of the IGO are reduced to a (discrete) Poincare map governing the propagation of the model's continuous states through the firing instants of the impulsive feedback. Bifurcation analysis of the map reveals in particular that both the local Hill function and affine feedback can lead to multistability, which phenomenon has not been observed in the usual IGO model.

In this paper, we propose variations of Willems' fundamental lemma that utilize second-order moments such as correlation functions in the time domain and power spectra in the frequency domain. We believe that using a formulation with estimated correlation coefficients is suitable for data compression, and possibly can reduce noise. Also, the formulations in the frequency domain can enable modeling of a system in a frequency region of interest.

This paper addresses the design of an active cyber-attack detection architecture based on multiplicative watermarking, allowing for detection of covert attacks. We propose an optimal design problem, relying on the so-called output-to-output l(2)-gain, which characterizes the maximum gain between the residual output of a detection scheme and some performance output. Although optimal, this control problem is non-convex. Hence, we propose an algorithm to design the watermarking filters by solving the problem suboptimally via LMIs. We show that, against covert attacks, the output-to-output l(2)-gain is unbounded without watermarking, and we provide a sufficient condition for boundedness in the presence of watermarks.

This paper addresses the issue of data injection attacks on control systems. We consider attacks which aim at maximizing system disruption while staying undetected in the finite horizon. The maximum possible disruption caused by such attacks is formulated as a non-convex optimization problem whose dual problem is a convex semi-definite program. We show that the duality gap is zero using S-lemma. To determine the optimal attack vector, we formulate a soft-constrained optimization problem using the Lagrangian dual function. The framework of dynamic programming for indefinite cost functions is used to solve the soft-constrained optimization problem and determine the attack vector. Using the Karush-Kuhn-Tucker conditions, we also provide necessary and sufficient conditions under which the obtained attack vector is optimal to the primal problem. Finally, we illustrate the results through numerical examples.

In this paper, we consider data driven control of Hammerstein systems. For such systems a common control structure is a transfer function followed by a static output nonlinearity that tries to cancel the input nonlinearity of the system, which is modeled as a polynomial or piece-wise linear function. The linear part of the controller is used to achieve desired disturbance rejection and tracking properties. To design a linear part of the controller, we propose a weighted average risk criterion with the risk being the average of the squared L2 tracking error. Here the average is with respect to the observations used in the controller and the weighting is with respect to how important it is to have good control for different impulse responses. This criterion corresponds to the average risk criterion leading to the Bayes estimator and we therefore call this approach Bayes control. By parametrizing the weighting function and estimating the corresponding hyperparameters we tune the weighting function to the information regarding the true impulse response contained in the data set available to the user for the control design. The numerical results show that the proposed methods result in stable controllers with performance comparable to the optimal controller, designed using the true input nonlinearity and true plant.

In this paper, we consider data driven control of Hammerstein systems. For such systems a common control structure is a transfer function followed by a static output nonlinearity that tries to cancel the input nonlinearity of the system, which is modeled as a polynomial or piece-wise linear function. The linear part of the controller is used to achieve desired disturbance rejection and tracking properties. To design a linear part of the controller, we propose a weighted average risk criterion with the risk being the average of the squared L2 tracking error. Here the average is with respect to the observations used in the controller and the weighting is with respect to how important it is to have good control for different impulse responses. This criterion corresponds to the average risk criterion leading to the Bayes estimator and we therefore call this approach Bayes control. By parametrizing the weighting function and estimating the corresponding hyperparameters we tune the weighting function to the information regarding the true impulse response contained in the data set available to the user for the control design. The numerical results show that the proposed methods result in stable controllers with performance comparable to the optimal controller, designed using the true input nonlinearity and true plant.

In this paper, we consider data driven control of Hammerstein systems. For such systems a common control structure is a transfer function followed by a static output nonlinearity that tries to cancel the input nonlinearity of the system, which is modeled as a polynomial or piece-wise linear function. The linear part of the controller is used to achieve desired disturbance rejection and tracking properties. To design a linear part of the controller, we propose a weighted average risk criterion with the risk being the average of the squared L2 tracking error. Here the average is with respect to the observations used in the controller and the weighting is with respect to how important it is to have good control for different impulse responses. This criterion corresponds to the average risk criterion leading to the Bayes estimator and we therefore call this approach Bayes control. By parametrizing the weighting function and estimating the corresponding hyperparameters we tune the weighting function to the information regarding the true impulse response contained in the data set available to the user for the control design. The numerical results show that the proposed methods result in stable controllers with performance comparable to the optimal controller, designed using the true input nonlinearity and true plant.

In this paper, we consider data driven control of Hammerstein systems. For such systems a common control structure is a transfer function followed by a static output nonlinearity that tries to cancel the input nonlinearity of the system, which is modeled as a polynomial or piece-wise linear function. The linear part of the controller is used to achieve desired disturbance rejection and tracking properties. To design a linear part of the controller, we propose a weighted average risk criterion with the risk being the average of the squared L2 tracking error. Here the average is with respect to the observations used in the controller and the weighting is with respect to how important it is to have good control for different impulse responses. This criterion corresponds to the average risk criterion leading to the Bayes estimator and we therefore call this approach Bayes control. By parametrizing the weighting function and estimating the corresponding hyperparameters we tune the weighting function to the information regarding the true impulse response contained in the data set available to the user for the control design. The numerical results show that the proposed methods result in stable controllers with performance comparable to the optimal controller, designed using the true input nonlinearity and true plant.

In this paper, we consider data driven control of Hammerstein systems. For such systems a common control structure is a transfer function followed by a static output nonlinearity that tries to cancel the input nonlinearity of the system, which is modeled as a polynomial or piece-wise linear function. The linear part of the controller is used to achieve desired disturbance rejection and tracking properties. To design a linear part of the controller, we propose a weighted average risk criterion with the risk being the average of the squared L2 tracking error. Here the average is with respect to the observations used in the controller and the weighting is with respect to how important it is to have good control for different impulse responses. This criterion corresponds to the average risk criterion leading to the Bayes estimator and we therefore call this approach Bayes control. By parametrizing the weighting function and estimating the corresponding hyperparameters we tune the weighting function to the information regarding the true impulse response contained in the data set available to the user for the control design. The numerical results show that the proposed methods result in stable controllers with performance comparable to the optimal controller, designed using the true input nonlinearity and true plant.

In this paper, we study the problem of detecting early signs of Parkinson's disease during an indirect human-computer interaction via a computer mouse activated by a user. The experimental setup provides a signal determined by the screen pointer position. An appropriate choice of segments in the cursor position raw data provides a filtered signal from which a number of quantifiable criteria can be obtained. These dynamical features are derived based on control theory methods. Thanks to these indicators, a subsequent analysis allows the detection of users with tremor. Real-life data from patients with Parkinson's and healthy controls are used to illustrate our detection method.

Several estimation methods have been proposed for identifying errors-in-variables systems, where both input and output measurements are corrupted by noise. One of the more interesting approaches is the Frisch scheme. The method can be applied using either time or frequency domain representations. This paper investigates the general mathematical and geometrical aspects of the Frisch scheme, illustrating the analogies and the differences between the time and frequency domain formulations.

This paper describes a new approach for the blind identification of a two-channel FIR system from a finite number of output measurements, in the presence of additive and uncorrelated white noise. The proposed approach is based on frequency domain data and, as a major novelty, it enables the estimation to be frequency selective. The features of the proposed method are analyzed by means of Monte Carlo simulations. The benefits of filtering the data and using only part of the frequency domain are highlighted by means of a numerical example.

Many physical quantities around us vary across space or space-time. An example of a spatial quantity is provided by the temperature across Sweden on a given day and as an example of a spatio-temporal quantity we observe the counts of the corona virus cases across the globe. Spatial and spatio-temporal data enable opportunities to answer many important questions. For example, what the weather would be like tomorrow or where the highest risk for occurrence of a disease is in the next few days? Answering questions such as these requires formulating and learning statistical models. One of the challenges with spatial and spatio-temporal data is that the size of data can be extremely large which makes learning a model computationally costly. There are several means of overcoming this problem by means of matrix manipulations and approximations. In paper I, we propose a solution to this problem where the model islearned in a streaming fashion, i.e., as the data arrives point by point. This also allows for efficient updating of the learned model based on newly arriving data which is very pertinent to spatio-temporal data. Another interesting problem in the spatial context is to study the causal effect that an exposure variable has on a response variable. For instance, policy makers might be interested in knowing whether increasing the number of police in a district has the desired effect of reducing crimes there. The challenge here is that of spatial confounding. A spatial map of the number of police against the spatial map of the number of crimes in different districts might show a clear association between these two quantities. However, there might be a third unobserved confounding variable that makes both quantities small and large together. In paper II, we propose a solution for estimating causal effects in the presence of such a confounding variable. Another common type of spatial data is point or event data, i.e., the occurrence of events across space. The event could f

Predicting unknown and unobserved events is a common task in many domains. Mathematically, the uncertainties arising in such prediction tasks can be described by probabilistic predictive models. Ideally, the model estimates of these uncertainties allow us to distinguish between uncertain and trustworthy predictions. This distinction is particularly important in safety-critical applications such as medical image analysis and autonomous driving. For the probabilistic predictions to be meaningful and to allow this differentiation, they should neither be over- nor underconfident. Models that satisfy this property are called calibrated. In this thesis we study how one can measure, estimate, and statistically reason about the calibration of probabilistic predictive models. In Paper I we discuss existing approaches for evaluating calibration in multi-class classification. We mention potential pitfalls and suggest hypothesis tests for the statistical analysis of model calibration. In Paper II we propose a framework of calibration measures for multi-class classification. It captures common existing measures and includes a new kernel calibration error based on matrix-valued kernels. For the kernel calibration error consistent and unbiased estimators exist and asymptotic hypothesis tests for calibration can be derived. Unfortunately, by construction the framework is limited to prediction problems with finite discrete target spaces. In Paper III we use a different approach to develop a more general framework of calibration errors that applies to any probabilistic predictive model and is not limited to classification. We show that it coincides with the framework presented in Paper II for multi-class classification. Based on scalar-valued kernels, we generalize the kernel calibration error, its estimators, and hypothesis tests to all probabilistic predictive models. For real-valued regression problems we present empirical results.

Predicting unknown and unobserved events is a common task in many domains. Mathematically, the uncertainties arising in such prediction tasks can be described by probabilistic predictive models. Ideally, the model estimates of these uncertainties allow us to distinguish between uncertain and trustworthy predictions. This distinction is particularly important in safety-critical applications such as medical image analysis and autonomous driving. For the probabilistic predictions to be meaningful and to allow this differentiation, they should neither be over- nor underconfident. Models that satisfy this property are called calibrated. In this thesis we study how one can measure, estimate, and statistically reason about the calibration of probabilistic predictive models. In Paper I we discuss existing approaches for evaluating calibration in multi-class classification. We mention potential pitfalls and suggest hypothesis tests for the statistical analysis of model calibration. In Paper II we propose a framework of calibration measures for multi-class classification. It captures common existing measures and includes a new kernel calibration error based on matrix-valued kernels. For the kernel calibration error consistent and unbiased estimators exist and asymptotic hypothesis tests for calibration can be derived. Unfortunately, by construction the framework is limited to prediction problems with finite discrete target spaces. In Paper III we use a different approach to develop a more general framework of calibration errors that applies to any probabilistic predictive model and is not limited to classification. We show that it coincides with the framework presented in Paper II for multi-class classification. Based on scalar-valued kernels, we generalize the kernel calibration error, its estimators, and hypothesis tests to all probabilistic predictive models. For real-valued regression problems we present empirical results.

Many physical quantities around us vary across space or space-time. An example of a spatial quantity is provided by the temperature across Sweden on a given day and as an example of a spatio-temporal quantity we observe the counts of the corona virus cases across the globe. Spatial and spatio-temporal data enable opportunities to answer many important questions. For example, what the weather would be like tomorrow or where the highest risk for occurrence of a disease is in the next few days? Answering questions such as these requires formulating and learning statistical models. One of the challenges with spatial and spatio-temporal data is that the size of data can be extremely large which makes learning a model computationally costly. There are several means of overcoming this problem by means of matrix manipulations and approximations. In paper I, we propose a solution to this problem where the model islearned in a streaming fashion, i.e., as the data arrives point by point. This also allows for efficient updating of the learned model based on newly arriving data which is very pertinent to spatio-temporal data. Another interesting problem in the spatial context is to study the causal effect that an exposure variable has on a response variable. For instance, policy makers might be interested in knowing whether increasing the number of police in a district has the desired effect of reducing crimes there. The challenge here is that of spatial confounding. A spatial map of the number of police against the spatial map of the number of crimes in different districts might show a clear association between these two quantities. However, there might be a third unobserved confounding variable that makes both quantities small and large together. In paper II, we propose a solution for estimating causal effects in the presence of such a confounding variable. Another common type of spatial data is point or event data, i.e., the occurrence of events across space. The event could f

Several estimation methods have been proposed for identifying errors-in-variables systems, where both input and output measurements are corrupted by noise. One of the more interesting approaches is the Frisch scheme. The method can be applied using either time or frequency domain representations. This paper investigates the general mathematical and geometrical aspects of the Frisch scheme, illustrating the analogies and the differences between the time and frequency domain formulations.

This paper describes a new approach for the blind identification of a two-channel FIR system from a finite number of output measurements, in the presence of additive and uncorrelated white noise. The proposed approach is based on frequency domain data and, as a major novelty, it enables the estimation to be frequency selective. The features of the proposed method are analyzed by means of Monte Carlo simulations. The benefits of filtering the data and using only part of the frequency domain are highlighted by means of a numerical example.