Given the large variety of existing logical formalisms it is of utmost importance to select the most adequate one for a specific purpose, e.g. for representing the knowledge relevant for a particular application or for using the formalism as a modeling tool for problem solving. Awareness of the nature of a logical formalism, in other words, of its fundamental intrinsic properties, is indispensable and provides the basis of an informed choice. One such intrinsic property of logic-based knowledge representation languages is the context-dependency of pieces of knowledge. In classical propositional logic, for example, there is no such context-dependence: whenever two sets of formulas are equivalent in the sense of having the same models (ordinary equivalence), then they are mutually replaceable in arbitrary contexts (strong equivalence). However, a large number of commonly used formalisms are not like classical logic which leads to a series of interesting developments. It turned out that sometimes, to characterize strong equivalence in formalism L, we can use ordinary equivalence in formalism L0: for example, strong equivalence in normal logic programs under stable models can be characterized by the standard semantics of the logic of here-and-there. Such results about the existence of characterizing logics has rightly been recognized as important for the study of concrete knowledge representation formalisms and raise a fundamental question: Does every formalism have one? In this thesis, we answer this question with a qualified “yes”. More precisely, we show that the important case of considering only finite knowledge bases guarantees the existence of a canonical characterizing formalism. Furthermore, we argue that those characterizing formalisms can be seen as classical, monotonic logics which are uniquely determined (up to isomorphism) regarding their model theory. The other main part of this thesis is devoted to argumentation semantics which play the flagship role in Dung’s abstract argumentation theory. Almost all of them are motivated by an easily understandable intuition of what should be acceptable in the light of conflicts. However, although these intuitions equip us with short and comprehensible formal definitions it turned out that their intrinsic properties such as existence and uniqueness, expressibility, replaceability and verifiability are not that easily accessible. We review the mentioned properties for almost all semantics available in the literature. In doing so we include two main axes: namely first, the distinction between extension-based and labelling-based versions and secondly, the distinction of different kind of argumentation frameworks such as finite or unrestricted ones.

To incorporate new technical advances into military domain and make those processes more efficient in accuracy, time and cost, a new concept of Network Centric Warfare has been introduced in the US military forces. In Sweden a similar concept has been studied under the name Network Based Defence (NBD). Here we present one of the methodologies, called tactical plan recognition that is aimed to support NBD in future. Advances in sensor technology and modelling produce large sets of data for decision makers. To achieve decision superiority, decision makers have to act agile with proper, adequate and relevant information (data aggregates) available. Information fusion is a process aimed to support decision makers’ situation awareness. This involves a process of combining data and information from disparate sources with prior information or knowledge to obtain an improved state estimate about an agent or phenomena. Plan recognition is the term given to the process of inferring an agent’s intentions from a set of actions and is intended to support decision making. The aim of this work has been to introduce a methodology where prior (empirical) knowledge (e.g. behaviour, environment and organization) is represented and combined with sensor data to recognize plans/behaviours of an agent or group of agents. We call this methodology multi-agent plan recognition. It includes knowledge representation as well as imprecise and statistical inference issues. Successful plan recognition in large scale systems is heavily dependent on the data that is supplied. Therefore we introduce a bridge between the plan recognition and sensor management where results of our plan recognition are reused to the control of, give focus of attention to, the sensors that are supposed to acquire most important/relevant information. Here we combine different theoretical methods (Bayesian Networks, Unified Modeling Language and Plan Recognition) and apply them for tactical military situations for ground forces. The results achieved from several proof-ofconcept models show that it is possible to model and recognize behaviour of tank units. / QC 20101222

<div>Federated learning (FL) is a decentralized privacy-preserving learning technique in which clients learn a joint collaborative model through a central aggregator without sharing their data. In this setting, all clients learn a single common predictor (FedAvg), which does not generalize well on each client's local data due to the statistical data heterogeneity among clients. In this paper, we address this problem with PersFL, a discrete two-stage personalized learning algorithm. In the first stage, PersFL finds the optimal teacher model of each client during the FL training phase. In the second stage, PersFL distills the useful knowledge from optimal teachers into each user's local model. The teacher model provides each client with some rich, high-level representation that a client can easily adapt to its local model, which overcomes the statistical heterogeneity present at different clients. We evaluate PersFL on CIFAR-10 and MNIST datasets using three data-splitting strategies to control the diversity between clients' data distributions.</div><div><br></div><div>We empirically show that PersFL outperforms FedAvg and three state-of-the-art personalization methods, pFedMe, Per-FedAvg and FedPer on majority data-splits with minimal communication cost. Further, we study the performance of PersFL on different distillation objectives, how this performance is affected by the equitable notion of fairness among clients, and the number of required communication rounds. We also build an evaluation framework with the following modules: Data Generator, Federated Model Generation, and Evaluation Metrics. We introduce new metrics for the domain of personalized FL, and split these metrics into two perspectives: Performance, and Fairness. We analyze the performance of all the personalized algorithms by applying these metrics to answer the following questions: Which personalization algorithm performs the best in terms of accuracy across all the users?, and Which personalization algorithm is the fairest amongst all of them? Finally, we make the code for this work available at https://tinyurl.com/1hp9ywfa for public use and validation.</div>

While methods like learning-without-forgetting [11] and elastic weight consolidation [22] accomplish high-quality transfer learning while mitigating catastrophic forgetting, progressive techniques such as Deepmind’s progressive neural network accomplish this while completely nullifying forgetting. However, progressive systems like this strictly require task labels during test time. In this paper, I introduce a novel task recognizer built from anomaly detection autoencoders that is capable of detecting the nature of the required task from input data.Alongside a progressive neural network or other progressive learning system, this task-aware network is capable of operating without task labels during run time while maintaining any catastrophic forgetting reduction measures implemented by the task model.

Intrusion Detection Systems are vital for computer networks as they protect against attacks that lead to privacy breaches and data leaks. Over the years, researchers have formulated intrusion detection systems (IDS) using machine learning and/or deep learning to detect network anomalies and identify four main attacks namely, Denial of Service (DoS), Probe, Remote to Local (R2L) and User to Root (U2R). However, the existing models are efficient in detecting just few of the aforementioned attacks while having inadequate detection rates for the rest. This deficiency makes it difficult to choose an appropriate IDS model when a user does not know what attacks to expect. Thus, there is a need for an IDS model that can detect, with uniform efficiency, all the four main classes of network intrusions. This research is aimed at exploring a machine learning approach to an intrusion detection model that can detect DoS, Probe, R2L and U2R attack classes with uniform and high efficiency. A multilayer perceptron was trained in an ensemble with J48 decision tree. The resultant ensemble learning model achieved over 85% detection rates for each of DoS, probe, R2L, and U2R attacks.

We provide a systematic in-depth study of the semantics of abstract dialectical frameworks (ADFs), a recent generalisation of Dung\''s abstract argumentation frameworks. This is done by associating with an ADF its characteristic one-step consequence operator and defining various semantics for ADFs as different fixpoints of this operator. We first show that several existing semantical notions are faithfully captured by our definition, then proceed to define new ADF semantics and show that they are proper generalisations of existing argumentation semantics from the literature. Most remarkably, this operator-based approach allows us to compare ADFs to related nonmonotonic formalisms like Dung argumentation frameworks and propositional logic programs. We use polynomial, faithful and modular translations to relate the formalisms, and our results show that both abstract argumentation frameworks and abstract dialectical frameworks are at most as expressive as propositional normal logic programs.

Abstract dialectical frameworks (ADFs) have recently been proposed as a versatile generalization of Dung''s abstract argumentation frameworks (AFs). In this paper, we present a comprehensive analysis of the computational complexity of ADFs. Our results show that while ADFs are one level up in the polynomial hierarchy compared to AFs, there is a useful subclass of ADFs which is as complex as AFs while arguably offering more modeling capacities. As a technical vehicle, we employ the approximation fixpoint theory of Denecker, Marek and Truszczyński, thus showing that it is also a useful tool for complexity analysis of operator-based semantics.

<p>Commercial buildings consume more than 19% of the total energy consumption in the United States. Most of this energy is consumed by the HVAC and shading/lighting systems inside these buildings. The main purpose of such systems is to provide satisfactory thermal and visual environments for occupants working inside these buildings. Providing satisfactory thermal/visual conditions in indoor environments is critical since it directly affects occupants’ comfort, health and productivity and has a significant effect on energy performance of the buildings. </p> <p>Therefore, efficiently learning occupants’ preferences is of prime importance to address the dual energy challenge of reducing energy usage and providing occupants with comfortable spaces at the same time. The objective of this thesis is to develop robust and easy to implement algorithms for learning and eliciting thermal and visual preferences of office occupants from limited data. As such, the questions studied in this thesis are: 1) How can we exploit concepts from utility theory to model (in a Bayesian manner) the hidden thermal and visual utility functions of different occupants? Our central hypothesis is that an occupant’s preference relation over different thermal/visual states of the room can be described using a scalar function of these states, which we call the “occupant’s thermal/visual utility function.” 2) By making use of formalisms in Bayesian decision theory, how can we learn the maximally preferred thermal/visual states for different occupants without requiring unnecessary or excessive efforts from occupants and/or the building engineers? The challenge here is to minimize the number of queries posed to the occupants to learn the maximally preferred thermal/visual states for each occupant. 3) Inferring preferences of occupants based on their responses to the thermal/visual comfort-based questionnaire surveys is intrusive and expensive. Contrary to this, how can we learn the thermal/visual preferences of occupants from cheap and non-intrusive human-building interactions’ data? 4) Lastly, based on the observation that the occupant population decompose into different clusters of occupants having similar preferences, how can we exploit the collective information obtained from the similarities in the occupants’ behavior? This thesis presents viable answers to the aforementioned questions in the form of probabilistic graphical models/frameworks. In future, I hope that these frameworks would prove to be an important step towards the development of intelligent thermal/visual systems which would be able to respond to occupants’ personalized comfort needs. Furthermore, in order to encourage the use of these frameworks and ensure reproducibility in results,various implementations of this work (namely GPPref, GPElicit and GPActToPref) are published as open-source Python packages.</p><br>

<p>The fast advancing of high-throughput technology has reinforced the biomedical research ecosystem with highly scaled and commercialized data acquisition standards, which provide us with unprecedented opportunity to interrogate biology in novel and creative ways. However, unraveling the high dimensional data in practice is difficult due to the following challenges: 1) how to handle outlier and data contaminations; 2) how to address the curse of dimensionality; 3) how to utilize occasionally provided auxiliary information such as an external phenotype observation or spatial coordinate; 4) how to derive the unknown non-linear relationship between observed data and underlying mechanisms in complex biological system such as human metabolic network. </p> <p><br></p> <p>In sight of the above challenges, this thesis majorly focused on two research directions, for which we have proposed a series of statistical learning and AI-empowered systems biology models. This thesis separates into two parts. The first part focuses on identifying latent low dimensional subspace in high dimensional biomedical data. Firstly, we proposed CAT method which is a robust mixture regression method to detect outliers and estimate parameter simultaneously. Then, we proposed CSMR method in studying the heterogeneous relationship between high dimensional genetic features and a phenotype with penalized mixture regression. At last, we proposed SRMR which investigate mixture linear relationship over spatial domain. The second part focuses on studying the non-linear relationship for human metabolic flux estimation in complex biological system.  We proposed the first method in this domain that can robustly estimate flux distribution of a metabolic network at the resolution of individual cells.</p>

The World Wide Web contains large amounts of data and in most cases this data is without any explicit structure. The lack of structure makes it difficult for automated agents to understand and use such data. A step towards a more structured World Wide Web is the idea of the Semantic Web which aims at introducing semantics to data on the World Wide Web. One of the key technologies in this endeavour are ontologies which provide means for modeling a domain of interest. Developing and maintaining ontologies is not an easy task and it is often the case that defects are introduced into ontologies. This can be a problem for semantically-enabled applications such as ontology-based querying. Defects in ontologies directly influence the quality of the results of such applications as correct results can be missed and wrong results can be returned. This thesis considers one type of defects in ontologies, namely the problem of completing the is-a structure in ontologies represented in description logics. We focus on two variants of description logics, the EL family and ALC, which are often used in practice. The contributions of this thesis are as follows. First, we formalize the problem of completing the is-a structure as a generalized TBox abduction problem (GTAP) which is a new type of abduction problem in description logics. Next, we provide algorithms for solving GTAP in the EL family and ALC description logics. Finally, we describe two implemented systems based on the introduced algorithms. The systems were evaluated in two experiments which have shown the usefulness of our approach. For example, in one experiment using ontologies from the Ontology Alignment Evaluation Initiative 58 and 94 detected missing is-a relations were repaired by adding 54 and 101 is-a relations, respectively, introducing new knowledge to the ontologies.