arXiv subjects
Publications and source records attributed to Marko Horvat.
In this work, we study the computability of topological graphs, which are obtained by gluing arcs and rays together at their endpoints. We prove that every semicomputable graph in a computable metric space can be approximated, with arbitrary precision, by its computable subgraph with computable endpoints.
This article presents a review of typical techniques used in three distinct aspects of deep learning model development for audio generation. In the first part of the article, we provide an explanation of audio representations, beginning with the fundamental audio waveform. We then progress to the frequency domain, with an emphasis on the attributes of human hearing, and finally introduce a relatively recent development. The main part of the article focuses on explaining basic and extended deep learning architecture variants, along with their practical applications in the field of audio generation. The following architectures are addressed: 1) Autoencoders 2) Generative adversarial networks 3) Normalizing flows 4) Transformer networks 5) Diffusion models. Lastly, we will examine four distinct evaluation metrics that are commonly employed in audio generation. This article aims to offer novice readers and beginners in the field a comprehensive understanding of the current state of the art in audio generation methods as well as relevant studies that can be explored for future research.
AMD Secure Encrypted Virtualization technologies enable confidential computing by protecting virtual machines from highly privileged software such as hypervisors. In this work, we develop the first, comprehensive symbolic model of the software interface of the latest SEV iteration called SEV Secure Nested Paging (SEV-SNP). Our model covers remote attestation, key derivation, page swap and live migration. We analyze the security of the software interface of SEV-SNP and formally prove that most critical secrecy, authentication, attestation and freshness properties do indeed hold in the model. Furthermore, we find that the platform-agnostic nature of messages exchanged between SNP guests and the AMD Secure Processor firmware presents a potential weakness in the design. We show how this weakness leads to formal attacks on multiple security properties, including the partial compromise of attestation report integrity, and discuss possible impacts and mitigations.
Reordering the terms of a series is a useful mathematical device, and much is known about when it can be done without affecting the convergence or the sum of the series. For example, if a series of real numbers absolutely converges, we can add the even-indexed and odd-indexed terms separately, or arrange the terms in an infinite two-dimensional table and first compute the sum of each column. The possibility of even more intricate re-orderings prompts us to find a general underlying principle. We identify such a principle in the setting of Banach spaces, where we consider well-ordered series with indices beyond ω, but strictly under ω_1 . We prove that for every absolutely convergent well-ordered series indexed by a countable ordinal, if the series is rearranged according to any countable ordinal, then the absolute convergence and the sum of the series remain unchanged.
Ever growing number of image documents available on the Internet continuously motivates research in better annotation models and more efficient retrieval methods. Formal knowledge representation of objects and events in pictures, their interaction as well as context complexity becomes no longer an option for a quality image repository, but a necessity. We present an ontology-based online image annotation tool WNtags and demonstrate its usefulness in several typical multimedia retrieval tasks using International Affective Picture System emotionally annotated image database. WNtags is built around WordNet lexical ontology but considers Suggested Upper Merged Ontology as the preferred labeling formalism. WNtags uses sets of weighted WordNet synsets as high-level image semantic descriptors and query matching is performed with word stemming and node distance metrics. We also elaborate our near future plans to expand image content description with induced affect as in stimuli for research of human emotion and attention.
Theoretical analysis of machine intelligence (MI) is useful for defining a common platform in both theoretical and applied artificial intelligence (AI). The goal of this paper is to set canonical definitions that can assist pragmatic research in both strong and weak AI. Described epistemological features of machine intelligence include relationship between intelligent behavior, intelligent and unintelligent machine characteristics, observable and unobservable entities and classification of intelligence. The paper also establishes algebraic definitions of efficiency and accuracy of MI tests as their quality measure. The last part of the paper addresses the learning process with respect to the traditional epistemology and the epistemology of MI described here. The proposed views on MI positively correlate to the Hegelian monistic epistemology and contribute towards amalgamating idealistic deliberations with the AI theory, particularly in a local frame of reference.
The Geneva Affective Picture Database WordNet Annotation Tool (GWAT) is a user-friendly web application for manual annotation of pictures in Geneva Affective Picture Database (GAPED) with WordNet. The annotation tool has an intuitive interface which can be efficiently used with very little technical training. A single picture may be labeled with many synsets allowing experts to describe semantics with different levels of detail. Noun, verb, adjective and adverb synsets can be keyword-searched and attached to a specific GAPED picture with their unique identification numbers. Changes are saved automatically in the tool's relational database. The attached synsets can be reviewed, changed or deleted later. Additionally, GAPED pictures may be browsed in the tool's user interface using simple commands where previously attached WordNet synsets are displayed alongside the pictures. Stored annotations can be exported from the tool's database to different data formats and used in 3rd party applications if needed. Since GAPED does not define keywords of individual pictures but only a general category of picture groups, GWAT represents a significant improvement towards development of comprehensive picture semantics. The tool was developed with open technologies WordNet API, Apache, PHP5 and MySQL. It is freely available for scientific and non-commercial use.
The ability to efficiently search pictures with annotated semantics and emotion is an important problem for Human-Computer Interaction with considerable interdisciplinary significance. Accuracy and speed of the multimedia retrieval process depends on the chosen metadata annotation model. The quality of such multifaceted retrieval is opposed to the potential complexity of data setup procedures and development of multimedia annotations. Additionally, a recent study has shown that databases of emotionally annotated multimedia are still being predominately searched manually which highlights the need to study this retrieval modality. To this regard we present a study with N = 75 participants aimed to evaluate the influence of keywords and dimensional emotions in manual retrieval of pictures. The study showed that if the multimedia database is comparatively small emotional annotations are sufficient to achieve a fast retrieval despite comparatively lesser overall accuracy. In a larger dataset semantic annotations became necessary for efficient retrieval although they contributed to a slower beginning of the search process. The experiment was performed in a controlled environment with a team of psychology experts. The results were statistically consistent with validates measures of the participants' perceptual speed.
Multimedia documents such as text, images, sounds or videos elicit emotional responses of different polarity and intensity in exposed human subjects. These stimuli are stored in affective multimedia databases. The problem of emotion processing is an important issue in Human-Computer Interaction and different interdisciplinary studies particularly those related to psychology and neuroscience. Accurate prediction of users' attention and emotion has many practical applications such as the development of affective computer interfaces, multifaceted search engines, video-on-demand, Internet communication and video games. To this regard we present results of a study with N=10 participants to investigate the capability of standardized affective multimedia databases in stimulation of emotion. Each participant was exposed to picture and video stimuli with previously determined semantics and emotion. During exposure participants' physiological signals were recorded and estimated for emotion in an off-line analysis. Participants reported their emotion states after each exposure session. The a posteriori and a priori emotion values were compared. The experiment showed, among other reported results, that carefully designed video sequences induce a stronger and more accurate emotional reaction than pictures. Individual participants' differences greatly influence the intensity and polarity of experienced emotion.
Multimedia documents such as images, sounds or videos can be used to elicit emotional responses in exposed human subjects. These stimuli are stored in affective multimedia databases and successfully used for a wide variety of research in affective computing, human-computer interaction and cognitive sciences. Affective multimedia databases are simple repositories of multimedia documents with annotated high-level semantics and affective content. Although important all affective multimedia databases have numerous deficiencies which impair their applicability. To establish a better understanding of how experts use affective multimedia databases an online survey was conducted into the subject. The survey results are statistically significant and indicate that contemporary databases lack stimuli with rich semantic and emotional content. 73.33% of survey participants find the databases lacking at least some important semantic or emotion content. Most of the participants consider stimuli descriptions to be inadequate. Overall, 1-2h or more than 24h are generally needed to construct a single stimulation sequence. Almost 84% of the survey participants would like to use real-life videos in their research. Experts unequivocally recognize the need for an intelligent stimuli retrieval application that would assist them in experimentation. Almost all experts agree such applications could be useful in their work.
Repositories of images with semantic and emotion content descriptions are valuable tools in many areas such as Affective Computing and Human-Computer Interaction, but they are also important in the development of multimodal searchable online databases. Ever growing number of image documents available on the Internet continuously motivates research of better annotation models and more efficient retrieval methods which use mash-up of available data on semantics, scenes, objects, events, context and emotion. Formal knowledge representation of such high-level semantics requires rich, explicit, human but also machine-processable information. To achieve these goals we present an online ontology-based image annotation tool WNtags and demonstrate its usefulness in knowledge representation and image retrieval using the International Affective Picture System database. The WNtags uses WordNet as image tagging glossary but considers Suggested Upper Merged Ontology as the preferred upper labeling formalism. The retrieval is performed using node distance metrics to establish semantic relatedness between a query and the collaboratively weighted tags describing high-level image semantics, after which the result is ranked according to the derived importance. We also elaborate plans to improve the WNtags to create a collaborative Web-based multimedia repository for research in human emotion and attention.
Affective multimedia documents such as images, sounds or videos elicit emotional responses in exposed human subjects. These stimuli are stored in affective multimedia databases and successfully used for a wide variety of research in psychology and neuroscience in areas related to attention and emotion processing. Although important all affective multimedia databases have numerous deficiencies which impair their applicability. These problems, which are brought forward in the paper, result in low recall and precision of multimedia stimuli retrieval which makes creating emotion elicitation procedures difficult and labor-intensive. To address these issues a new core ontology STIMONT is introduced. The STIMONT is written in OWL-DL formalism and extends W3C EmotionML format with an expressive and formal representation of affective concepts, high-level semantics, stimuli document metadata and the elicited physiology. The advantages of ontology in description of affective multimedia stimuli are demonstrated in a document retrieval experiment and compared against contemporary keyword-based querying methods. Also, a software tool Intelligent Stimulus Generator for retrieval of affective multimedia and construction of stimuli sequences is presented.
Administration of a Web directory and maintenance of its content and the associated structure is a delicate and labor intensive task performed exclusively by human domain experts. Subsequently there is an imminent risk of a directory structures becoming unbalanced, uneven and difficult to use to all except for a few users proficient with the particular Web directory and its domain. These problems emphasize the need to establish two important issues: i) generic and objective measures of Web directories structure quality, and ii) mechanism for fully automated development of a Web directory's structure. In this paper we demonstrate how to formally and fully integrate Web directories with the Semantic Web vision. We propose a set of criteria for evaluation of a Web directory's structure quality. Some criterion functions are based on heuristics while others require the application of ontologies. We also suggest an ontology-based algorithm for construction of Web directories. By using ontologies to describe the semantics of Web resources and Web directories' categories it is possible to define algorithms that can build or rearrange the structure of a Web directory. Assessment procedures can provide feedback and help steer the ontology-based construction process. The issues raised in the article can be equally applied to new and existing Web directories.
The administration of a Web directory content and associated structure is a labor intensive task performed by human domain experts. Because of that there always exists a realistic risk of the structure becoming unbalanced, uneven and difficult to use to all except for a few users proficient in a particular Web directory. These problems emphasize the importance of generic and objective measures of Web directories structure quality. In this paper we demonstrate how to formally merge Web directories into the Semantic Web vision. We introduce a set of objective criterions for evaluation of a Web directory's structure quality. Some criteria functions are based on heuristics while others require the application of ontologies.
Emotionally annotated databases are repositories of multimedia documents with annotated affective content that elicit emotional responses in exposed human subjects. They are primarily used in research of human emotions, attention and development of stress-related mental disorders. This can be successfully exploited in larger processes like selection, evaluation and training of personnel for occupations involving high stress levels. Emotionally annotated databases are also used in multimodal affective user interfaces to facilitate richer and more intuitive human-computer interaction. Multimedia documents in emotionally annotated databases must have maximum personal ego relevance to be the most effective in all these applications. For this reason flexible construction of subject-specific of emotionally annotated databases is imperative. But current construction process is lengthy and labor intensive because it inherently includes an elaborate tagging experiment involving a team of human experts. This is unacceptable since the creation of new databases or modification of the existing ones becomes slow and difficult. We identify a positive correlation between the affect and semantics in the existing emotionally annotated databases and propose to exploit this feature with an interactive relevance feedback for a more efficient construction of emotionally annotated databases. Automatic estimation of affective annotations from existing semantics enhanced with information refinement processes may lead to an efficient construction of high-quality emotionally annotated databases.
For over 50 years, astronomers have searched the skies for evidence of electromagnetic signals from extraterrestrial civilizations that have reached or surpassed our level of technological development. Although often overlooked or given as granted, the parallel use of an equivalent communication technology is a necessary prerequisite for establishing contact in both leakage and deliberate messaging strategies. Civilization advancements, especially accelerating change and exponential growth, lessen the perspective for a simultaneous technological status of civilizations thus putting hard constraints on the likelihood of a dialogue. In this paper we consider the mathematical probability of technological synchronicity of our own and a number of other hypothetical extraterrestrial civilizations and explore the most likely scenarios for their concurrency. If SETI projects rely on a fortuitous detection of leaked interstellar signals (so called "eavesdropping") then without any prior assumptions N \geq 138-4991 Earth-like civilizations have to exist at this moment in the Galaxy for the technological usage synchronicity probability p \geq 0.95 in the next 20 years. We also show that since the emergence of complex life, coherent with the hypothesis of the Galactic habitable zone, N \geq 1497 extraterrestrial civilizations had to be created in the Galaxy in order to achieve the same estimated probability in the technological possession synchronicity which corresponds to the deliberate signaling scenario.
Successful management of emotional stimuli is a pivotal issue concerning Affective Computing (AC) and the related research. As a subfield of Artificial Intelligence, AC is concerned not only with the design of computer systems and the accompanying hardware that can recognize, interpret, and process human emotions, but also with the development of systems that can trigger human emotional response in an ordered and controlled manner. This requires the maximum attainable precision and efficiency in the extraction of data from emotionally annotated databases While these databases do use keywords or tags for description of the semantic content, they do not provide either the necessary flexibility or leverage needed to efficiently extract the pertinent emotional content. Therefore, to this extent we propose an introduction of ontologies as a new paradigm for description of emotionally annotated data. The ability to select and sequence data based on their semantic attributes is vital for any study involving metadata, semantics and ontological sorting like the Semantic Web or the Social Semantic Desktop, and the approach described in the paper facilitates reuse in these areas as well.
Although it might not be self-evident, it is in fact entirely possible to calculate the probability of detecting alien radio signals by understanding what types of extraterrestrial radio emissions can be expected and what properties these emissions can have. Using the Drake equation as the obvious starting point, and logically identifying and enumerating constraints of interstellar radio communications can yield the probability of detecting a genuine alien radio signal.