Academic literature on the topic 'Neural-Symbolic Artificial Intelligence MaxEntropy Neural Networks General AI'

Abstract Smart building and smart city specialists agree that complex, innovative use cases, especially those using cross-domain and multi-source data, need to make use of Artificial Intelligence (AI). However, today’s AI mainly concerns machine learning and artificial neural networks (deep learning), whereas the first forty years of the discipline (the last decades of the 20th century) were essentially focused on a knowledge-based approach, which is still relevant today for some tasks. In this article we advocate a merging of these two AI trends – an approach known as neuro-symbolic AI – for the smart city, and point the way towards a complete integration of the two technologies, compatible with standard software.

The need for neural-symbolic integration becomes apparent as more complex problems are tackled, and they go beyond limited domain tasks such as classification. In this sense, understanding the state of the art of hybrid technologies based on Deep Learning and augmented with logic based systems, is of utmost importance. As a consequence, we seek to understand and represent the current state of these technologies that are highly used in intelligent systems engineering.This work aims to provide a comprehensive view of the solutions available in the literature, within the field of applied Artificial Intelligence (AI), using technologies based on AI techniques that integrate symbolic and non-symbolic logic (in particular artificial neural networks), making them the subject of a systematic literature review (SLR). The resulting technologies are discussed and evaluated from both perspectives: symbolic and non-symbolic AI.In this work, we use the PICOC & Limits method to define the research questions and analyze the results.Out of a total of 65 candidate studies found, 24 articles (37%) relevant to this study were selected. Each study also focuses on different application domains. Conclusion: Through the analysis of the selected works throughout this review, we have seen different combinations of logical systems with some form of neural network and, although we have not found a clear architectural pattern, efforts to find a model of general purpose combining both worlds drive trends and research efforts.

During the past decade, several engineering disciplines, including construction, have embarked on developing “intelligent” decision support systems based on artificial intelligence (AI) techniques, including expert systems, symbolic knowledge representation, and logic programming. These systems attempt to capture the domain experts' intelligent behaviour and reasoning process utilized in decision-making, without regard to the underlying mechanisms producing that behaviour. This approach involves describing behaviours, usually with rules and symbols. In contrast, neural networks (NN), another AI-based technique that has been pursued on a large scale during the past few years, does not describe behaviours but rather imitate them. Neural networks are particularly superior to traditional expert systems in providing timely solutions based primarily on analogy with previous experience, rather than reasoning or computation. As such, neural networks have a great potential to work either as a supplement or as a complement to algorithmic and (or) other AI-based systems, providing more suitable tools for solving the industry ill-structured problems.This paper describes several characteristics of neural networks and outlines the advantages and limitations of commonly used NN paradigms. Potential applications of each paradigm in construction are identified. Two example applications are provided to demonstrate the problem-solving capabilities of neural networks: (i) estimation of hourly production rate of an excavation equipment; and (ii) estimation of productivity level for a construction trade. Future possibilities of integrating neural networks with other problem-solving techniques are described. Key words: construction, management techniques, neural networks, expert systems, pattern recognition, computer applications.

Learning and reasoning are both aspects of what is considered to be intelligence. Their studies within AI have been separated historically, learning being the topic of machine learning and neural networks, and reasoning falling under classical (or symbolic) AI. However, learning and reasoning are in many ways interdependent. This paper discusses the nature of some of these interdependencies and proposes a general framework called FLARE, that combines inductive learning using prior knowledge together with reasoning in a propositional setting. Several examples that test the framework are presented, including classical induction, many important reasoning protocols and two simple expert systems.

Abstract Artificial Intelligence (AI) systems based solely on neural networks or symbolic computation present a representational complexity challenge. While minimal representations can produce behavioral outputs like locomotion or simple decision-making, more elaborate internal representations might offer a richer variety of behaviors. We propose that these issues can be addressed with a computational approach we call meta-brain models. Meta-brain models are embodied hybrid models that include layered components featuring varying degrees of representational complexity. We will propose combinations of layers composed using specialized types of models. Rather than using a generic black box approach to unify each component, this relationship mimics systems like the neocortical-thalamic system relationship of the mammalian brain, which utilizes both feedforward and feedback connectivity to facilitate functional communication. Importantly, the relationship between layers can be made anatomically explicit. This allows for structural specificity that can be incorporated into the model's function in interesting ways. We will propose several types of layers that might be functionally integrated into agents that perform unique types of tasks, from agents that simultaneously perform morphogenesis and perception, to agents that undergo morphogenesis and the acquisition of conceptual representations simultaneously. Our approach to meta-brain models involves creating models with different degrees of representational complexity, creating a layered meta-architecture that mimics the structural and functional heterogeneity of biological brains, and an input/output methodology flexible enough to accommodate cognitive functions, social interactions, and adaptive behaviors more generally. We will conclude by proposing next steps in the development of this flexible and open-source approach.

Over the past ten years, molecular biologists and computer scientists have experimented with various computational methods developed in artificial intelligence (AI). AI research has yielded a number of novel technologies, which are typified by an emphasis on symbolic (non-numerical) programming methods aimed at problems which are not amenable to classical algorithmic solutions. Prominent examples include knowledge-based and expert systems, qualitative simulation and artificial neural networks and other automated learning techniques. These methods have been applied to problems in data analysis, construction of advanced databases and modelling of biological systems. Practical results are now being obtained, notably in the recognition of active genes in genomic sequences, the assembly of physical and genetic maps and protein structure prediction. This paper outlines the principal methods, surveys the findings to date, and identifies the promising trends and current limitations.

The current urban design computation is mostly centered on the professional designer while ignoring the plural dimension of urban design. In addition, available public participation computational tools focus mainly on information and idea sharing, leaving the public excluded in design generation because of their lack of design expertise. To address such an issue, this study develops Urban-GAN, a plural urban design computation system, to provide new technical support for design empowerment, allowing the public to generate their own designs. The sub-symbolic representation and artificial intelligence techniques of deep convolutional neural networks, case-based reasoning, and generative adversarial networks are used to acquire and embody design knowledge as the density function, and generate design schemes with this knowledge. The system consists of an urban form database and five process models through which the user with little design expertise can select urban form cases, generate designs similar to those cases, and make design decisions. The Urban-GAN is applied to hypothetical design experiments, which show that the user is able to apply the system to successfully generate distinctive designs following the urban form “styles” in Manhattan, Portland, and Shanghai. This study further extends the discussion about the plural urban design computation to general reflections on the goals and values in AI technique application in planning and design.

Deep Learning is considered the foundation of a new technological revolution, which is going to radically change human beings lives. No doubt that, in the following years, the use of deep learning techniques will bring to the development of outstanding applications that will totally change the way people interact with each other, interact with their environment, conceive work, travel, etc. However, while in the domain of applications deep learning is destined to revolutionize any human activity, it is extremely difficult to believe that these techniques, in an isolated way, can lead to the great dream of general artificial intelligence. Indeed, deep learning algorithms have shown to reach human-like performance in many isolated tasks, like image recognition or speech recognition, but they still struggle in integrating these single activities in a truly intelligent behaviour. Moreover, they are tremendously data-hungry: is recognizing a cat after having seen one million cats an intelligent behaviour? There is a clear need to look forward for more complex and general theories, where the outstanding deep learning techniques are not a final recipe but only an ingredient. We hope this thesis to be a little step in this direction. By taking inspiration from both symbolic artificial intelligence, strongly based on mathematical logics, and behavioural sciences, this thesis formulates and investigates a new theory about how intelligent behaviours can be the outcome of the seamlessly integration of two reasoning mechanisms: one subsymbolic, associative and fast; the other symbolic, cautious and slow. This integration naturally emerges from a principle of MiniMax Entropy, where the model of an intelligent system is asked to describe the environment it is exposed to by minimizing its internal confusion, yet keeping the maximum uncertainty about anything it is not able to explain. The theory nicely intercepts multiple AI fields like Statistical Relational Learning, Deep Learning, Constrained Optimization, Probabilistic Graphical Models, Neuro-Symbolic Integration, to name a few. The theory is extremely general and multiple models can be described in terms of MiniMax Entropy models. Some practical instances of the theory are proposed and empirically investigated, showing competitive results in many different learning and reasoning tasks w.r.t. other specific state-of-the-art approaches. Moreover, actual programming frameworks, inspired by the theory, are proposed and provided to the community for future investigations.

Knowledge-based artificial intelligence has been extensively used in numerous application areas leading to the development of a vast number of methods and tools. In recent years, focus has shifted on non-symbolic approaches, and neural networks in particular have achieved human-level performance in various applications where accountability is a very important issue, closely related to the interpretability of artificial intelligence methods in general. Lack of interpretability of neural networks and various machine learning methods has led to the adoption of knowledge-based methods instead, which offer models compliant with explainability and interpretability requirements. In this article, an overview of knowledge-based methods is presented along with the state of the art in this area, offering to the AI practitioner guidance for applying these important methods in practice.

A cognitive system is any real system that has some cognitive property. Therefore, cognitive systems are a special type of K-systems (see chapter 3, section 3). Note that this definition includes both natural systems such as humans and other animals, and artificial devices such as robots, implementations of AI (artificial intelligence) programs, some implementations of neural networks, etc. Focusing on what all cognitive systems have in common, we can state a very general but nonetheless interesting thesis: All cognitive systems are dynamical systems. Section 2 explains what this thesis means and why it is (relatively) uncontroversial. It will become clear that this thesis is a basic methodological assumption that underlies practically all current research in cognitive science. The goal of section 3 is to contrast two styles of scientific explanation of cognition: computational and dynamical. Computational explanations are characterized by the use of concepts drawn from computation theory, while dynamical explanations employ the conceptual apparatus of dynamical systems theory. Further, I will suggest that all scientific explanations of cognition might end up sharing the same dynamical style, for dynamical systems theory may well turn out to be useful in the study of all types of models currently employed in cognitive science. In particular, a dynamical viewpoint might even benefit those scientific explanations of cognition which are based on symbolic models. Computational explanations of cognition, by contrast, can only be based on symbolic models or, more generally, on any other type of computational model. In particular, those scientific explanations of cognition which are based on an important class of connectionist models cannot be computational, for this class of models falls beyond the scope of computation theory. Arguing for this negative conclusion requires the formal explication of the concept of a computational system that I gave in chapter 1 (see definition 3). Finally, section 4 explores the possibility that scientific explanations of cognition might be based on Galilean models of cognitive systems (see chapter 3, section 5). Most cognitive scientists have not yet considered this possibility. The goals of this section are to contrast this proposal with the current modeling practice in cognitive science, to make clear its potential benefits, and to indicate possible ways to implement it.

Neural-symbolic computing has now become the subject of interest of both academic and industry research laboratories. Graph Neural Networks (GNNs) have been widely used in relational and symbolic domains, with widespread application of GNNs in combinatorial optimization, constraint satisfaction, relational reasoning and other scientific domains. The need for improved explainability, interpretability and trust of AI systems in general demands principled methodologies, as suggested by neural-symbolic computing. In this paper, we review the state-of-the-art on the use of GNNs as a model of neural-symbolic computing. This includes the application of GNNs in several domains as well as their relationship to current developments in neural-symbolic computing.