Search

Recent advances in autonomous vehicles promise to revolutionize the transportation system. This perspective has led to new research on a number of open questions, such as how the self-driving system should behave in unavoidable crash situations. Our study aims to contribute to this investigation. In most ongoing research, this question is presented as a moral dilemma, drawing on established research on the trolley dilemma. However, more recent studies have shifted the focus from morality to risk analysis. We investigated the dual contribution of moral judgment and risk analysis in subjects facing dangerous situations. To this end, we use virtual reality to recreate a driving situation that allows subjects to immerse themselves in the road environment. Our results show a strong dissociation between quick decisions, when risk analysis seems to suggest the best choice, and conscious decisions, when proper moral judgment is at stake.

This paper analyzes the rapid and unexpected rise of deep learning within Artificial Intelligence and its applications. It tackles the possible reasons for this remarkable success, providing candidate paths towards a satisfactory explanation of why it works so well, at least in some domains. A historical account is given for the ups and downs, which have characterized neural networks research and its evolution from “shallow” to “deep” learning architectures. A precise account of “success” is given, in order to sieve out aspects pertaining to marketing or sociology of research, and the remaining aspects seem to certify a genuine value of deep learning, calling for explanation. The alleged two main propelling factors for deep learning, namely computing hardware performance and neuroscience findings, are scrutinized, and evaluated as relevant but insufficient for a comprehensive explanation. We review various attempts that have been made to provide mathematical foundations able to justify the efficiency of deep learning, and we deem this is the most promising road to follow, even if the current achievements are too scattered and relevant for very limited classes of deep neural models. The authors’ take is that most of what can explain the very nature of why deep learning works at all and even very well across so many domains of application is still to be understood and further research, which addresses the theoretical foundation of artificial learning, is still very much needed.

Cyber Therapy is a research project based on the relationship between Computer Science and Psychology. We are working on post-traumatic stress disorder (PTSD) due to severe traffic accidents using a virtual reality driving simulator and ECG, EDA and breathing sensors. With the help of virtual reality (VR) our goal is to build one software that can process the user’s biofeedback signals - heart rate, body temperature, state of tension, etc. - in real time, to make the phobic stimulus autonomous. To this purpose, we developed a platform capable of adapting the phobic stimulus based on the user’s biofeedback signals. We believe that this human-computer integrated system could be useful to patients as it would allow them to face fear autonomously, and to the psychotherapist, as it would allow a real time - physiologically based - knowledge of fear symptoms severity able to promote a timely and more appropriate program of intervention.

Autonomous driving has, in recent years, gained considerable traction due to commercial interest by car manufacturers, that envision a widespread use of self-driving cars in the next decade. Research has mainly focused on specific topics, that can enable the feasibility of robust artificial autonomous agents for autonomous vehicles (AVs), ranging from artificial vision, to automatic environmental sensing and inference, agent-agent interaction and sensor augmentation and integration. A somewhat reduced effort has been directed towards the evaluation of human-AV interaction and in the direction of defining acceptable rules of engagement when AV deal with human drivers. This aspect of autonomous driving will became extremely relevant in the next years as level 5 AVs will make their appearance on our streets and most of the cars will still be driven by humans. To assess, experimentally, how AVs can successfully interact with human drivers we have constructed a lab setup, which can simulate a large range of situations in which human drivers encounter AVs and interact with them. Based on the Unity Gaming Engine an urban environment has been developed where AVs interact with human driven cars, pedestrians and various agents. This setup allows for human subjects to drive in a virtual reality (VR) environment, and assess their behavior during the interaction with other human drivers and AVs. The goal is to define a set of rules that can be applied in AV design, to make autonomous cars react to human drivers, taking into account the behavioral patterns and variable respect of traffic regulations that human drivers exhibit in real environments. We present an early experimental setup, potentially useful for understanding, in a quantitative and reproducible way, how this approach can contribute in designing AVs that are more suited for large scale deployment.

Our study aims at progressing the assessment of the moral behaviour of human drivers. It is a felicitous coincidence that psychological and philosophical research into human morality has been dominated by thought experiments, resembling vehicles facing emergency situations. These thought experiments involve a running trolley, and have been used to contrast different moral principles, especially deontology versus utilitarianism. We designed an ecologically valid trolley-like dilemma with the help of virtual reality, aimed to understand the moral behavior of human subjects when facing a car accident situation. We report and comment on early results of our first tests.

Business analytics use techniques from data science, data mining, artificial intelligence (especially, machine learning), mathematics and statistics to gain insights and understanding on the performance of business processes. The gained insights and knowledge help driving the business planning. As employees play important roles in the business process, having a tool to classify and predict their wage levels is desirable. Such classification and prediction enables the public or private sector to offer competitive wages for recruiting and retaining employees. In this paper, we present a tool for classifying and predicting wage levels. It incorporates fuzzy logic into a machine-learning tool to support business analytics on big data. Evaluation results show the applicability of our tool for classification and prediction of wages levels in the business world, which in turn supports business analytics in complex artificial intelligence environments.

In this paper, we provide principles, models, and main architecture of an innovative framework for supporting intelligent analytics over big atmospheric data via clustering-based spatio-temporal analysis. In particular we investigates the interesting applicative setting represented by Greenhouse Gas Emissions (GGEs), a relevant instance of Big Data that empathize the Variety aspect of the well-known 3V Big Data axioms. A relevant case study is also introduced and discussed in detail. We also provide a comprehensive experimental evaluation of the proposed framework, which indeed confirms the benefits of our approach. The deriving Big Data Mining model turns to be useful for decision support processes in both the governmental and industrial contexts. We complete our analytical contributions by means of concluding remarks of our work, and a vision on future research efforts in the field.

In this paper, we describe AppxDL, an algorithm for approximate classification of workloads of running processes in big data environments via deep learning (deep neural networks). The Deep Neural Network is trained with some workloads which belong to known categories (e.g., compiler, file compressor, etc...). Its purpose is to extract the type of workload from the executions of reference programs, so that a Neural Model of the workloads can be learned. When the learning phase is completed, the Deep Neural Network is available as Neural Model of the known workloads. We describe the AppxDL algorithm and we report and discuss some significant results we have achieved with it.

Recent advances in Artificial Intelligence are making automated vehicles an ever closer reality. However, we should expect a period when full or partial autonomous vehicles and ordinary cars coexist, during which it would be essential to fully understand the cognitive processes used by ordinary people when driving. Our work attempt to progress in this direction, by designing a system for assessing when and why subjects resort to costly social processes, rather than using quick and automated reactions. In particular, it will be crucial to assess when drivers use mentalizing abilities, in addition to paying attention to other people by means of simpler automated sensorimotor control processes. In our experimental design we investigate the main precursors of mindreading, that is, eye contact and shared attention.

Recent developments in smart assistants and smart home automation are lately attracting the interest and curiosity of consumers and researchers. Speech enabled virtual assistants (often named smart speakers) offer a wide variety of network-oriented services and, in some cases, can connect to smart environments, thus enhancing them with new and effective user interfaces. However, such devices also reveal new needs and some weaknesses. In particular, they represent faceless and blind assistants, unable to show a face, and therefore an emotion, and unable to ‘see’ the user. As a consequence, the interaction is impaired and, in some cases, ineffective. Moreover, most of those devices heavily rely on cloud-based services, thus transmitting potentially sensitive data to remote servers. To overcome such issues, in this paper we combine some of the most advanced techniques in computer vision, deep learning, speech generation and recognition, and artificial intelligence, into a virtual assistant architecture for smart home automation systems. The proposed assistant is effective and resource-efficient, interactive and customizable, and the realized prototype runs on a low-cost, small-sized, Raspberry PI 3 device. For testing purposes, the system was integrated with an open source home automation environment and ran for several days, while people were encouraged to interact with it, and proved to be accurate, reliable and appealing.

In the current era of big data, high volumes of a wide variety of valuable data of different veracity are generated or collected at a high velocity. A rich source of these big data is the stock market. Since the inception of the stock market, people have been trying to "beat" it for the purpose of monetary gain. A stock market is an exchange where people trade shares of companies, also called stocks. The purpose of the exchange is to make it easy to match buyers and sellers together to make transactions. The usual goal of someone participating in the stock market it to generate profit through the buying and selling of stocks. The main way people accomplish this is by buying a stock, waiting anywhere from seconds to decades, and then hopefully selling for more than they bought it for. This is where the common term "buy low, sell high" comes from. There are many factors (e.g., hurricanes) that may affect the stock price. In this paper, we present a computational intelligent tool that applies fuzzy logic-based data analytics to predict the effect of hurricanes on the stock market.

This paper describes a novel distributing computing middleware named XML-VM. Its architecture is inspired by the ‘Grid Computing’ paradigm. The proposed system improves many characteristics of previous Grid systems, in particular the description of the distributed computation, the distribution of the code and the execution times. XML is a markup language commonly used to interchange arbitrary data over the Internet. The idea behind this work is to use XML to describe algorithms; XML documents are distributed by means of XML-RPC, interpreted and executed using virtual machines. XML-VM is an assembly-like language, coded in XML. Parsing of XML-VM programs is performed with a fast SAX parser for JAVA. XML-VM interpreter is coded in JAVA. Several algorithms are written in XML-VM and executed in a distributed environment. Representative experimental results are reported.

In the current era of big data, high volumes of a wide variety of data of different veracity can be easily generated or collected at a high velocity from rich sources of data include devices from the Internet of Things (IoT). Embedded in these big data are useful information and valuable knowledge. Hence, frequent pattern mining and its related research problem of association rule mining, which aim to discover implicit, previously unknown and potentially useful information and knowledge—in the form of sets of frequently co-occurring items or rules revealing relationships between these frequent sets—from these big data have drawn attention of many researchers. For instance, since introduction of the research problems of association rule mining or frequent pattern mining, numerous information system and engineering approaches have been developed. These include the development of serial algorithms, distributed and parallel algorithms, as well as MapReduce-based big data mining algorithms. These algorithms can be run in local computers, distributed and parallel environments, as well as clusters, grids and clouds. In this paper, we describe some of these algorithms and discuss how to mine frequent patterns or association rules in fogs—i.e., edges of the computing network.

Color re-indexing is one of main approaches for improving the loss-less compression of color indexed images. Zero-order entropy reduction of indexes matrix is the key to obtain high compression ratio. However, obtaining the optimal re-indexed palette is a challenging problem that cannot be solved by brute-force approaches. In this paper we propose a novel re-indexing approach where the Travelling Salesman Problem is solved through Ant Colony Optimization. Our method is proved to achieve high quality results by outperforming state-of-art ones in term of compression gain. Additionally, we exploit clustering and GPU computing to make our solution extremely fast.

This paper proposes applying and experimentally assessing machine learning tools to solve security issues in complex environments, specifically identifying and analyzing malicious behaviors. To evaluate the effectiveness of machine learning algorithms to detect anomalies, we consider the following three real-world case studies: (i) detecting and analyzing Tor traffic, on the basis of a machine learning-based discrimination technique; (ii) identifying and analyzing CAN bus attacks via deep learning; (iii) detecting and analyzing mobile malware, with particular regard to ransomware in Android environments, by means of structural entropy-based classification. Derived observations confirm the effectiveness of machine learning in supporting security of complex environments.

This paper attempts to analyze the theoretical consequences for conceptual integrity in software engineering, if concepts do not exist altogether. As weird as it might sound, doubts on the existence of concepts have been recently cast by several distinguished philosophers of mind, with compelling arguments. It is commonly asserted that one of the most crucial factors in software design is the adoption of consistent and appropriate concepts. This widespread assumption, underlying several researches in software engineering, tacitly postulates the very existence of concepts. How the integrity of a software design may survive without concepts, at least in its traditional account, is the matter explored in this paper. We identified several practical cases of software that uses design ideas referred by terms subsuming a heterogeneity of cognitive assets. These design ideas are therefore better qualified — in technical terms — as unicepts. This notwithstanding, in the cases analyzed, all software maintained a design integrity that made them highly successful, once users become acquainted with the unusual semantic sphere of the unicept at play.

Dynamic programming is a popular optimization technique, developed in the 60’s and still widely used today in several fields for its ability to find global optimum. Dynamic Programming Algorithms (DPAs) can be developed in many dimension. However, it is known that if the DPA dimension is greater or equal to two, the algorithm is an NP complete problem. In this paper we present an approximation of the fully two-dimensional DPA (2D-DPA) with polynomial complexity. Then, we describe an implementation of the algorithm on a recent parallel device based on CUDA architecture. We show that our parallel implementation presents a speed-up of about 25 with respect to a sequential implementation on an Intel I7 CPU. In particular, our system allows a speed of about ten 2D-DPA executions per second for 85 × 85 pixels images. Experiments and case studies support our thesis.

This paper provides an overview of some state-ofthe- art proposals in the context of querying encrypted OLAP data, along with critical discussion on open challenges and future research directions in the investigated topics.

In this paper, we describe an algorithm to estimate the parameters of Iterated Function System (IFS) fractal models. We use IFS to model Speech and Electroencephalographic signals and compare the results. The IFS parameters estimation is performed by means of a genetic optimization approach. We show that the estimation algorithm has a very good convergence to the global minimum. This can be successfully exploited by pattern recognition tools. However, the set-up of the genetic algorithm should be properly tuned. In this paper, besides the optimal set-up description, we describe also the best tradeoff between performance and computational complexity. To simplify the optimization problem some constraints are introduced. A comparison with suboptimal algorithms is reported. The performance of IFS modeling of the considered signals are in accordance with known measures of the fractal dimension.