ABSTRACT. The hard problem of consciousness is the problem of explaining how we experience qualia or phenomenal experiences, such as seeing, hearing, and feeling, and knowing what they are. To solve this     problem, a theory of consciousness needs to link brain to mind by modeling how brain mechanisms give rise to conscious psychological experiences, notably how emergent properties of several brain mechanisms interacting together embody parametric properties of conscious psychological experiences. This lecture summarizes evidence that Adaptive Resonance Theory, or ART, accomplishes this goal. ART is currently the most advanced cognitive and neural theory, with the broadest explanatory and predictive range, of how advanced brains autonomously learn to attend, recognize, and predict objects and events in a changing world. In particular, ART can incrementally learn, fast or slow, unsupervised    or supervised, in a self-stabilizing way in response to a complex non-stationary world filled with unexpected events. It also provides functional and mechanistic explanations of data ranging from individual spikes and their synchronization to the dynamics of conscious perceptual, cognitive, and    cognitive-emotional experiences.

ART has predicted that "all conscious states are resonant states" as part of its specification of mechanistic links between processes of consciousness, learning, expectation, attention, resonance, and synchrony. The theory has reached sufficient maturity to begin classifying the brain resonances that support conscious experiences of seeing, hearing, feeling, and knowing. Psychological and neurobiological data about conscious experiences in normal individuals and clinical patients are clarified by this classification. This analysis also explains why not all resonances are conscious, and why not all brain dynamics are resonant. Two revolutionary computational paradigms that describe different aspects of brain computing figure prominently in these explanations of conscious and unconscious processing; namely, Complementary Computing: the global organization of the brain into computationally complementary pairs of cortical processing streams, whose interactions overcome computational deficiencies of individual cortical streams; and Laminar Computing: the organization of the cerebral cortex into laminar circuits whose specializations can support multiple types of biological    intelligence, to date including laminar models of vision, audition, and cognition.

ABSTRACT. The paper discusses a viable approach to endow a cognitive architecture with artificial creativity to make a humanoid able to dance. Robot combines movements and music perception to create an aesthetically valuable dance by a suitable Hidden Markov Model. The model is mainly influenced by two matrices: Transition matrix TM, and Emission Matrix EM. EM constitutes the link between movements and music model, and we propose a genetic approach to define it under teacher evaluation, and determining different styles of dances (i.e. robot’s repertoire). We achieve different robot dancing tendencies using different fitness functions. Moreover we include in our analysis, a self evaluating fitness function that allows the robot teach itself a music-movements association.

ABSTRACT. We present a robot aimed at producing a collage formed by a mix of photomontage and digital collage. The artwork is created after a visual and verbal interaction with an human user. A cognitive architecture allows the robot to manage the three different phases of the real-time artwork process: (i) taking inspiration from postural and verbal interaction with the human user, including the exploration and the analysis of his/her social web items; (ii) performing a creative thinking process to obtain a mental model of the artwork; (iii) executing a creative collage composition and deciding a significant title. The paper explains, primarly, how creativity traits of the robot are influenced by the proposed architecture: how ideas are generated through cognitive thinking and affective influences; how the personality and the artistic behavior are modeled by learning and evaluations; the motivation, and the confidence evolution as a function of successes or failures.

ABSTRACT. The problem of interpreting the aesthetic emotions (induced by no pragmatic goal, but impression of artwork, Nature phenomena, etc.) is considered within the Natural-Constructive Approach to mod-eling the cognitive process. This approach is based on the dynamical theory of information, neuro-physiology data, and neural computing. The cognitive architecture designed under this approach repre-sents complex multi-level combination of various-type neural processors, with the whole system being divided into two subsystems, in analogy with two hemispheres of human brain. A peculiar feature of the architecture is the fuzzy set at the lowest (raw-image reception) hierarchy level. This processor contains the images recorded by weak (“grey”) connections that reflect individual (but not strictly formulated) experience. According to our hypothesis, these “grey” images are responsible for personal aesthetic preferences. The concept of masterpiece is associated with the “paradox of recognition”, which arises due to ambiguous impression (familiar and unexpected simultaneously) induced by the artwork.

ABSTRACT. Dynamical unstable processes in the brain: a biologically inspired communication mechanism from "conscious" to "unconscious" actors.

ABSTRACT. An autonomous actor should decide on its own which goals and strategies to pursue in a new situation involving multiple actors. Humans in such cases typically rely on social factors, such as individual relationships and ethical background. An artificial autonomous agent in such cases can be more useful and efficient as an actor in a human team, if its behavior is believable, i.e., similar to the naturally motivated human behavior. This similarity can be achieved in a cognitive architecture through the attribution of characters to actors and human-like reasoning in terms of ethical norms and moral schemas applied to developing individual relationships among characters. Whether the actor's behavior is sufficiently human-like and human-compatible in this sense, can be judged based on a Turing-like test that is described and analyzed here in simplistic videogame settings. The challenge for an artificial actor is to be preferred, over its human rival, as a trustworthy partner of the human participant. Additional metrics include behavioral characteristics derived from the study of cognitive architecture eBICA. The paradigm extends to other settings as well, and can be useful for evaluation of cognitive architectures that support near-human-level social emotionality.

ABSTRACT. In this paper, we propose a new Convolutional Neural Network (CNN) with biologically inspired retinal structure and ON/OFF Rectified Linear Unit (ON/OFF ReLU). Retinal structure enhances input images by center surround difference of green-red and blue-yellow components, which in turn creates positive as well as negative features like ON/OFF visual pathway of retina to make a total of 12 feature channels. This ON/OFF concept is also adopted to each convolutional layer of CNN. We prefer to call this ON/OFF ReLU. In contrast, conventional ReLU passes only positive features of each convolutional layer and may loose important information from negative features. Moreover, it also happens to loose learning chance if results are saturated to zero. However, in our proposed model, we use both positive and negative information, which provides a possibility to learn through negative results. We also present the experimental results conducted on CIFAR-10 dataset and atrial fibrillation prediction for health monitoring, and show how effectively the negative information and retinal structure improves the performance of conventional CNN.

ABSTRACT. The current research paper introduces the basic principles of the Information Images theory and mathematical model created through it. In order to confirm the theory experimentally, the bilingual Stroop test was used. The results of the test are interpreted through the introduced theory, then they are compared with the results of computer modeling on its basis. The authors demonstrate, that through the Information Images Theory it is possible not only to explain a number of cognitive processes of human mind, but also to make a prognosis of their dynamics in a number of isolated incidents.

ABSTRACT. The article considers the impact of modern Web communications on the psychophysical characteristics of youth. The authors conducted an experimental study of the influence of modern virtual-communication tools on the basic psychological and psychophysical characteristics of schoolchildren (12-13 years old) in terms of the Information Images Theory. It was found that such parameters as trait anxiety, emotional stress and excitability of the right hemisphere is largely subject to changes depending on incoming information and its intensity.

ABSTRACT. The problem of representing and learning complex visual stimuli in the context of modeling the process of conditional reflex formation is considered. The generative probabilistic framework is chosen which has been recently successfully applied to cognitive modeling. A model capable of learning different visual stimuli is developed in the form of a program in Church (probabilistic programming language). NAO robot is programmed to detect visual stimuli, to point at selected stimuli in a sequence of trials, and to receive reinforcement signals for correct choices. Conducted experiments showed that the robot can learn stimuli of different types showing different decision-making behavior in a series of trial that could help arranging psychophysiological experiments.

ABSTRACT. Classification of motor imagery electroencephalogram (EEG) is one of the most important technologies for BCI. To improve the accuracy, this paper introduces a classification system based on Multilayer Extreme Learning Machine (ML-ELM). In the system, the combination of PCA and LDA is chosen as the method of feature extraction and the ML-ELM is used to classify. The ML-ELM has not only the advantage which ELM has but also better performance than ELM. In the experiment, our method is compared with the methods based on ELM, such as kernel-ELM, Constrained-ELM and V-ELM, and some state-of–the-art methods on the same dataset. The experimental results show that ML-ELM is much more suitable for motor imagery EEG data and has better performance than the others.

ABSTRACT. Autism Spectrum Disorders (ASD) is a lifelong communication disorder that limits the abilities of diagnosed individuals to relate socially and interpret emotional cues. Thus, it is important to have early interventions in the domains of social and emotion or affective functioning for such individuals. Recent research efforts have focused on the innovative applications of Assistive Technologies (AT) for rehabilitation efforts. However, despite excellent preliminary findings, the efficacy of AT remains limited. This paper aims to fill the identified efficacy gap by proposing a framework incorporating music as a therapy which will be developed into a technological application to help children with autism to deal with their emotional dysfunctions. The proposal is also based on findings which show that this special population prefers and has successfully used technological devices such as the iPad for learning new skills.

ABSTRACT. A cognitive extension for a behavior-based control system for a six-legged robot is proposed which allows the robot to deal with novel situations. The underlying Walknet model is biologically inspired by detailed studies on walking stick insects and consists of a decentralized architecture from which stable and adaptive walking emerges. Following a minimal cognitive systems approach this system is extended towards a system capable of planning ahead which utilizes a functional internal model of its own body in mental simulation. While the body model is grounded in the underlying control system it is also capable of prediction and allows therefore for internal simulation. In this paper, the process of internal simulation is described in detail. It is detailed using the example of leg loss in a six-legged robot. First, it is explained how behaviors are selected for mental simulation and than how they are applied in the body model.

ABSTRACT. Cognitive phenomenology refers to the idea that our subjective experiences include deliberative thought processes and high-level cognition. The recent ascendance of cognitive phenomenology in philosophy has important implications for biologically-inspired cognitive architectures and the role that these models can play in understanding the fundamental nature of consciousness. To the extent that cognitive phenomenology occurs, it provides a new route to a deeper understanding of consciousness via neurocomputational studies of cognition. This route involves identifying computational correlates of consciousness in neurocomputational models of high-level cognitive functions that are associated with subjective mental states. Here we develop this idea and compile a summary of potential neurocomputational correlates of consciousness that have been proposed/recognized during the last several years based on biologically-inspired cognitive architectures. We conclude that the identification and study of computational correlates of consciousness will lead to a better understanding of phenomenal consciousness, a framework for creating a conscious machine, and a better understanding of the mind-brain problem in general.

ABSTRACT. During the last decades, many cognitive architectures (CAs) have been realized adopting different as- sumptions about the organization and the representation of their knowledge level. Some of them (e.g. SOAR) adopt a classical symbolic approach, some (e.g. LEABRA) are based on a purely con- nectionist models, while others (e.g. CLARION) adopt a hybrid approach combining connectionist and symbolic representational levels. Additionally some attempts (e.g. biSOAR) trying to embed di- agrammatical representation and reasoning in CAs are also available. In this paper we propose a reflection on the role that Conceptual Spaces, a framework developed by Peter Ga ̈rdenfors [20] more than fifteen years ago, can play in the current development of Knowledge Level in Cognitive Systems and Architectures. In particular, we claim that Conceptual Spaces offer a lingua franca that allows to unify and generalize many aspects of the symbolic, sub-symbolic and diagrammatic approaches (by overcoming the typical problems of such representations) and to integrate them on a common ground. In doing so we extend and detail some of the arguments explored by Gardenfors for defending the need of a conceptual, intermediate, level between the symbolic and the sub-symbolic one. In particular we focus on the advantages offered by Conceptual Spaces (w.r.t. symbolic and sub-symbolic approaches) in dealing with the problem of compositionally based on typicality traits. Additionally we argue that Conceptual Spaces could offer a unifying framework for interpreting many kinds of diagrammatic and analogical representations.

ABSTRACT. In evolutionary terms, imagery developed hundreds of millions of years before symbolic or language-like systems of cognition. Even the most abstract reasoning in science and mathematics requires imagery:  diagrams and written symbols supplement short-term memory, and richer imagery is essential for novel analogies and creative insights. A cognitive architecture must relate symbols to the perceptions and purposive actions of an embodied mind that interacts with the world and with other minds in it. This article reviews the evidence for an internal virtual reality as the foundation for the perception, action, and cognition of an embodied mind. Peirce’s theory of signs is a unifying framework that relates all branches of cognitive science, including AI implementations. The result is a theory of virtual reality for cognitive architectures (VRCA) that spans the minds from fish to humans and perhaps beyond.

ABSTRACT. In the recent years, Deep Learning has proved to be extremely effective in areas like Computer Vision, Natural Language Processing and Voice Recognition. Further enhancement of this technology has been made possible with the advent of Neuromorphic Computing. It is of utmost importance that designing of deep learning models be studied and understood to be able to make the most of this technology. In this paper, we propose a Deep Neural Networks (DNN) Design Index which would aid a DNN designer during the designing phase of DNNs. We study the designing aspect of DNNs from model-specific and data-specific perspectives with focus on three performance metrics: training time, training error and, validation error. We use a simple example to illustrate the significance of the DNN design index. To validate it, we calculate the design indices for four benchmark problems. This is an elementary work aimed at setting a direction for creating design indices pertaining to deep learning.

ABSTRACT. We present Knowledge Of Action Networks, which provide an enactive machine learning model for *knowledge of agency* in artificial intelligence. These networks, which are expected to be part of embodied intelligences existing in dynamic environments, learn to represent their environment while simultaneously learning to represent their own actions and bodies within that environment. Thus self and world are intricately coupled in their basic representations.

We will also explore some of the (many) expected contributions of such networks for implementing *minimal self-models*, which are basic models of self-aware agents. We finally describe an ongoing research program to integrate these networks into Google DeepMind's Atari-playing system.

ABSTRACT. The authors have successfully developed a consciousness system capable of autonomously increasing its knowledge by simply connecting that additional knowledge to the system. In this approach, the system always learns anew, and does not utilize the relearning method typical of conventional neural networks. Experiments proved that this consciousness system provides almost the same color vision capability as a human.

ABSTRACT. This study proposes a conscious model for a robot that achieves self-evolution by evaluating whether the internal state of the conscious system is represented as pleasant or unpleasant. For the definition of pleasant and unpleasant, we adopted the "smoothness of the flow of information" in the conscious system. Thus, our proposed conscious robot is capable of self-evolving tailored to the environment in which the robot itself is placed. In other words, it is possible to identify whether or not any external information has been input, and to identify the information with time series memory.

ABSTRACT. We conducted a simulation experiment focused on the development mechanism of self-consciousness using a neural network structured by consciousness modules that we had developed called Module of Nerves for Advanced Dynamics (MoNAD), and succeeded in presenting a system relevant to the development of self-cognition and self-consciousness. This system can suppress, through learning, “imitation behavior as a result of the cognition of another” in response to a “behavior of avoiding pain after feeling the pain by oneself,” and can “formulate” the cognition of oneself. These study results may be highly suggestive of the development of self-consciousness. In this paper, we describe the theoretical basis for the development of the self-cognition of our system, and discuss the experiment and observations.

ABSTRACT. This study proposes a consciousness model that simulates the onset process of the symptoms of Dissociative Identity Disorder (DID) using a conscious system constructed with Module of Nerves for Advanced Dynamics (MoNAD) consciousness modules. This paper discusses the possibility of employing the conscious system under development in order to promote a better understanding of severe human metal disorders such as Post Traumatic Stress Disorder (PTSD) and DID. We considered the symptoms of DID to be a problem of what is called “self-dispersion” in robot science, and developed a program with which we believe it will become possible to grasp the development process of DID more objectively.

ABSTRACT. Many studies have demonstrated how cognitive psychology can be used to elucidate human information processing processes such as memory, sensation and perception. This study takes up a classical cognitive problem in Gestalt psychology known as Rubin`s Vase. A very simple classical figure-ground organization problem, Rubin`s Vase is an ambiguous image that was developed by the Danish psychologist Edgar Rubin around 1915. The image appears to the viewer as two black faces looking at each other on a white background, or as a white vase on a black background. The image can be interpreted as only one of these two at a time. Rubin`s Vase has been the topic of previous research, and many mysteries still remain regarding psychological determinants, neurophysiological mechanisms and so on. Believing that Rubin`s Vase can be simulated by information science and robotics, the authors aimed to explain the phenomenon using a conscious system, and demonstrate how to construct a model using an artificial consciousness module they have developed called the Module of Nerves for Advanced Dynamics (MoNAD). Through the study of Rubin`s Vase, the authors conclude with the belief that many complex perceptions of humans can be simulated using neural networks, and that this can help us to study in depth the cognitive processes of human perception.

ABSTRACT. This study discusses signal propagation time delay effects in large neuronal nets. We use an approximated mathematical model of the highly nonlinear equations describing both artificial and biological neuronal nets in the presence of large propagation delays. We investigate the effect of delays on the computational capability and complexity of these nets, and their effect of on the neuronal net energy efficiency. We found that the presence of large time delays in biological neuronal nets may actually increase the number of available states to represent knowledge and improve their parallelism. Another effect we demonstrated is that it is impossible to have a simultaneous determination of both state values and their derivatives in time. This effect may enhance the computational capability of the nets at the price of decreasing state observability. We investigate how time delays effects may have influenced the evolution of biological neuronal nets.