Connectionist Networks

Bohm's holistic reframing of physics in Wholeness and the Implicate Order (1980) has direct relevance for consciousness studies. Bohm's interest in meaning reflects his theory of wholeness, with its structure of implicate and the explicate orders, which, together, form the 'holomovement'. In this paper, we expand on Bohm's wholeness approach to science and his linking of meaning to mind to produce a coherent wholeness model of consciousness. This model goes beyond traditional dualist approaches to philosophy, mind and consciousness by privileging the agency of the mental and does not rely on the causality of any physical functions or structures, such as the brain or nervous system. The model's two integrated levels-universal consciousness and local minds-are based on the distinctions of implicit and explicit meaning, the grounds of Bohm's implicate and explicate orders. We call this organic totality, with its absolute fullness of meaning, Omni-local consciousness.

We present a general discussion concerning the wholeness of what has been called infinite awareness, but here is called Omni-local consciousness. This model of consciousness has an interconnecting structure that has both local and nonlocal features, that is, the model contains local conscious human minds and locates them within an infinite (Omni) background context of consciousness. This holistic model of Omni-local consciousness is exemplified through an examination of its internal structures of meaning, evident in the exchange relations between its two polarities: local minds and nonlocal, Omni consciousness. Following David Bohm's assertion that, 'The activity of consciousness is determined by meaning' [10, p. 102], we propose that the content of consciousness in every circumstance is always defined by the metaphysical conditions of meaning.

Defined by enigmatic phenomena such as superposition, uncertainty, and entanglement, the quantum world represents the foundational reality beneath the perceivably deterministic classical environment. Despite our experiential interaction with the tangible classical world, it is crucial to acknowledge the elusive nature of the underlying quantum framework. This paper explores the ramifications of these quantum characteristics in management science.

The task allocation problem in a distributed environment is one of the most challenging problems in a multiagent system. We propose a new task allocation process using deep reinforcement learning that allows cooperating agents to act automatically and learn how to communicate with other neighboring agents to allocate tasks and share resources. Through learning capabilities, agents will be able to reason conveniently, generate an appropriate policy and make a good decision. Our experiments show that it is possible to allocate tasks using deep Q-learning and more importantly show the performance of our distributed task allocation approach.

In this paper, we argue that Bohm's unbroken and undivided totality he called the holomovement, the title he gave to the concept of the self-organizing universe, is more coherently understood when viewed as universal consciousness. Bohm's understanding of consciousness oscillates around being a quality of local minds and the interconnected totality of the holomovement. We suggest such equivocations impose limitations on Bohm's general holistic framework because they import into his model the limiting restrictions of Cartesian separation and are, therefore, incongruous for use within his holistic model of the holomovement. We also argue that the term 'meaning' has a structural and functional agency appropriate to Bohm's model of the holomovement, while also reflecting the living characteristics of this organic totality that is full of meaning.

Existing traffic light controls are ineffective and causes a handful of problems such as congestion and pollution. The purpose of this study is to investigate the application of deep reinforcement learning on traffic control systems to minimize congestion at a targeted traffic intersection. The traffic data was extracted, analyzed and simulated based on the Poisson Distribution, using a simulator, Simulation of Urban Mobility (SUMO). In this research, we proposed a deep reinforcement learning model, which combines the capabilities of convolutional neural networks and reinforcement learning to control the traffic lights to increase the effectiveness of the traffic control system. The paper explains the method we used to quantify the traffic scenario into different matrices which fed to the model as states which reduces the load of computing as compared to images. After 2000 iterations of training, our deep reinforcement learning model was able to reduce the cumulative waiting time of...

In this essay I will attempt to define the dream and the experience of free will in terms of models of holographic processing in the brain. There are two basic models with similar results. Accordingly, in the first, Schrodinger (but not quantum physical) wave holography is predicted to occur through active transport of Na+ (sodium) and P+ (potassium) ions in glial tissue in brain cortex. In the second model a temporal analogue of optical holography is employed in imaging so that cortical cells act as resonators much like a band of strings in a piano. Memory functions in the brain by making and witnessing holographic "imagery" and this functioning depends to a great deal on "who" is looking as well as on "what" is being observed. In these models we find in waking reality partial sensorial input evokes a complete, but possibly noisy, picture of what was previously sensed. Inner knowing of a volitional self is assumed already present. In dreams "out t...

Recent research in task transfer and task clustering has necessitated the need for task similarity measures in reinforcement learning. Determining task similarity is necessary for selective transfer where only information from relevant tasks and portions of a task are transferred. Which task similarity measure to use is not immediately obvious. It can be shown that no single task similarity measure is uniformly superior. The optimal task similarity measure is dependent upon the task transfer method being employed. We define similarity in terms of tasks, and propose several possible task similarity measures, dT , dP , dQ, and dR which are based on the transfer time, policy overlap, Q-values, and reward structure respectively. We evaluate their performance in three separate experimental situations.

The spontaneous self-organization of complex physical structures is a central issue in artificial intelligence. If self-organization were indeed a natural process, then 'cognition', i.e. all the mental processes that involve perception, memory, reasoning and learning, could be reduced to simple information processing, to more or less complex calculations carried out by the neuronal networks of our brain. This would naturally open the door to the emergence of an artificial, inhuman cognition, based on infinitely powerful computational structures over which we would eventually lose all control.

Historically, models of dreams and dreaming have been situated either in the physiology or the psychology of the dreamer, that is, the physical or psychological state of the dreamer is generally thought to be the ground from which dream imagery springs, either arbitrarily or by design. Few models, however, take contemporary models of the universe into account in describing the function and process of dreaming. This model of dreaming, which I will call the holo-net model, attempts to situate the dreamer and her dreams within Bohm and Pribram’s holographical models of the universe and brain.

The holo-net model is rooted in the connectionist net theory but differs in three essential ways: (a) the network of available imagery extends into the holomovement of the universal hologram, (b) dream images are highly individual, and (c) there is two-way communication between the consciousness of the dreamer and the holomovement.

Historically, models of dreams and dreaming have been situated either in the physiology or the psychology of the dreamer, that is, the physical or psychological state of the dreamer is generally thought to be the ground from which dream imagery springs, either arbitrarily or by design. Few models, however, take contemporary models of the universe into account in describing the function and process of dreaming. This model of dreaming, which I will call the holo-net model, attempts to situate the dreamer and her dreams within Bohm and Pribram’s holographical models of the universe and brain.

The holo-net model is rooted in the connectionist net theory, but differs from it in three essential ways: 1. the network of available imagery extends into the holomovement of the universal hologram, 2. dream images are highly individual, and 3. there is two-way communication between the consciousness of the dreamer and the holomovement.

The current dominant thinking about the nature of living things and the cognition with which they are endowed with is that their functionalities must all be reduced to « algorithms ». That is, sets of operating rules, instructions, applying to the development of more or less complex physico-chemical interactions such as oxygen and hydrogen gases that combine to form water. We would thus be highly sophisticated machines which would have spontaneously self-constructed by having the physically paradoxical capacity, given the entropic degradation of the universe, to ensure the durability of their structure in all circumstances, what characterizes them. This purely computational approach to life is not, however, scientifically founded.

Self-learning robots are physically unfeasible. Are the computational theories of cognition founded ? Michel Troublé – director of research, robotics and artificial intelligence. Robot autonomy is a major issue. The aim is to create physicochemical structures that are artificially alive and, as such, endowed with capacities similar to those of living beings in terms of learning and recognition the objects of the world with which structures interact. Robots are autonomous if they receive no human assistance regarding the choice of possible solutions for the actions they must perform to ensure the durability of their structure and related functionalities. As such, the cognition of a robot is what makes it capable of self-learning and self-recognizing the various objects of the world from the continuously changing information provided by its sensors. To be autonomous, a robot, a sensorimotor system, must have at least the following elements : a sensor which links the structure with the external environment, an actuator which provides the motility of the structure, a controller that must establish appropriate information links between the sensor and the actuator. The controller, as a sequential or parallel computing structure like formal neural network), establishes the cognition of the robot, that is to say the mechanisms that determine the actions to be taken for the robot to be autonomous. To ensure the durability of an autonomous robot, the functional analysis shows that for each object of the world, with infinitely varied properties, that it perceives with the help of its various sensors, it is necessary that its controller (its 'brain') know how to properly choose the actions to be performed by means of its actuator (locomotion system). For example, a land exploration robot sensitive to any temperature rise that can destroy it must, to be artificially alive, imperatively perform the following actions : run away from molten lava, a forest fire, an oil slick inflamed, ... In other words, the controller of this autonomous robot must be able to self-create a coherent category of objects perceived by its sensors, in other words the category {run away all hot objects}, while there is no physical affinity between the controller and the perceived objects.-The capability to create coherent categories of actions of perceived objects is an operational definition of cognition whose primary role is to ensure the durability of the autonomous physicochemical structure, artificial or living, with which it is associated – 'Theorem of indistinguishability of objects' From the point of view of the decision-making mechanism with which the robot controller must be equipped, this capability to create coherent categories logically implies that the different objects perceived by the system are distinguishable from each other. Otherwise these decisions will be taken at random which is clearly antinomic from the expected capability to form coherent categories on which the autonomy of the robot is founded. That the objects of the world perceived by the robot sensors are physically distinguishable seems obvious. But in fact it is a major epistemological problem. Based on the formal theory of 'Pattern Recognition' 1 , which is concerning with the identification of the shape of objects from their characteristic parameters in order to make decisions depending on the categories assigned to these forms, we can prove the following essential feature :-The different shapes of objects that are perceived by a physical system (or physico-chemical) during a measurement process, are physically indistinguishable by its operative part or actuator.

ABSTRACT Robust autonomous agents should be able to cooperate with new teammates effectively by employing ad hoc teamwork. Reasoning about ad hoc teamwork allows agents to perform joint tasks while cooperating with a variety of teammates. As the teammates may not share a communication or coordination algorithm, the ad hoc team agent adapts to its teammates just by observing them.

Recently researchers have introduced methods to develop reusable knowledge in reinforcement learning (RL). In this paper, we define simple principles to combine skills in reinforcement learning. We present a skill combination method that uses trained skills to solve different tasks in a RL domain. Through this combination method, composite skills can be used to express tasks at a high level and they can also be re-used with different tasks in the context of the same problem domains. The method generates an abstract task representation based upon normal reinforcement learning which decreases the information coupling of states thus improving an agent's learning. The experimental results demonstrate that the skills combination method can effectively reduce the learning space, and so accelerate the learning speed of the RL agent. We also show in the examples that different tasks can be solved by combining simple reusable skills.