21,821 research outputs found
A modeling environment for reified temporal-causal network models
Full text linkThe introduced multilevel reified (temporal-causal) network architecture is the basis of the implementation of a dedicated software environment developed by the author in Matlab. The environment includes a combination function library and a generic computational reified network engine. It uses role matrices specifying the characteristics for the designed network model as input. Based on this input, the computational reified network engine can be used to generate simulations for the network model, thereby using combination functions from the library. In this chapter, this software environment is described in more detail.</p
With a Little Help: A Modeling Environment for Self-modeling Network Models
Full text linkThe concept of self-modeling network architecture is the basis of the implementation of a dedicated software environment developed by the author in MATLAB. The environment includes a combination function library and a generic computational self-modeling network engine. It uses as input role matrices specifying in a standardised and compact form the characteristics for the designed network model. Based on this input, the computational self-modeling network engine can be used to generate simulations for the network model, thereby using combination functions from the library. In this chapter, this software environment and how to use it is described in more detail.</p
On the universal combination function and the universal difference equation for reified temporal-causal network models
Full text linkThe universal differential and difference equation form an important basis for reified temporal-causal networks and their implementation. In this chapter, a more in depth analysis is presented of the universal differential and difference equation. It is shown how these equations can be derived in a direct manner and they are illustrated by some examples. Due to the existence of these universal difference and differential equation, the class of temporal-causal networks is closed under reification: by them it can be guaranteed that any reification of a temporal-causal network is itself also a temporal-causal network. That means that dedicated modeling and analysis methods for temporal-causal networks can also be applied to reified temporal-causal networks. In particular, it guarantees that reification can be done iteratively in order to obtain multilevel reified network models that are very useful to model multiple orders of adaptation. Moreover, as shown in Chap. 9, the universal difference equation enables that software of a very compact form can be developed, as all reification levels are handled by one computational reified network engine in the same manner. Alternatively, it is shown how the universal difference or differential equation can be used for compilation by multiple substitution for all states, which leads to another form of implementation. The background of these issues is discussed in the current chapter.</p
A unified approach to represent network adaptation principles by network reification
Full text linkIn this chapter, the notion of network reification is introduced: a construction by which a given (base) network is extended by adding explicit states representing the characteristics defining the base network’s structure. This is explained for temporal-causal networks where connection weights, combination functions, and speed factors represent the characteristics for Connectivity, Aggregation, and Timing describing the network structure. Having the network structure represented in an explicit manner within the extended network enables to model the adaptation of the base network by dynamics within the reified network: an adaptive network is represented by a non-adaptive network. It is shown how the approach provides a unified modeling perspective on representing network adaptation principles across different domains. This is illustrated for a number of well-known network adaptation principles such as for Hebbian learning in Mental Networks and for network evolution based on homophily in Social Networks.</p
Using multilevel network reification to model second-order adaptive bonding by homophily
Full text linkThe concept of multilevel network reification introduced in the previous chapters enables representation within a network not only of first-order adaptation principles, but also of second-order adaptation principles expressing change of characteristics of first-order adaptation principles. In the current chapter, this approach is illustrated for an adaptive Social Network. This involves a first-order adaptation principle for bonding by homophily represented at the first reification level, and a second-order adaptation principle describing change of characteristics of this first-order adaptation principle, and represented at the second reification level. The second-order adaptation addresses adaptive change of two of the characteristics of the first-order adaptation, specifically similarity tipping point and connection adaptation speed factor.</p
Using network reification for adaptive networks:Discussion
Full text linkIn this final chapter, the most important or most remarkable themes recurring at different places in this book are briefly summarized and reviewed. Subsequently the following themes are addressed: (1) How network reification can be used to model adaptive networks. (2) The formats in which conceptual representations of reified networks are expressed graphically using 3D pictures and role matrices. (3) The universal combination function, and the universal difference and differential equation as the basis for the numerical representation and implementation of reified networks. (4) Analysis of how emerging reified network behaviour relates to the reified network’s structure. (5) The Network-Oriented design process based on reified networks. (6) The relation to longstanding themes in AI and beyond.</p
Modeling higher-order adaptive evolutionary processes by reified adaptive network models
Full text linkIn this chapter, a fourth-order reified network model is introduced to describe different orders of adaptivity found in a case study on evolutionary processes. The network model describes how the causal pathways for newly developed features in this case study affect the causal pathways of already existing features, which makes the pathways of these new features one order of adaptivity higher than the existing ones, as they adapt a previous adaptation. The network reification approach is shown to be an adequate means to model this transparently.</p
On adaptive networks and network reification
Full text linkThis chapter is a brief preview of what can be expected in this book, with some pointers to various chapters and sections. First, it is discussed how networks can be adaptive in different ways and according to different orders. A variety of examples of first and second-order adaptation are summarized, and the possibility of adaptation of order higher than two is discussed. After this, the notion of network reification is briefly summarized and how it can be used to model adaptive networks in a transparent and network-oriented manner. It is pointed out how repeated application of network reification can be used to model adaptive networks with the adaptation of multiple orders. Finally, it is discussed how mathematical analysis of emerging behavior of a network not only can be applied to non-adaptive base networks, but also to reified adaptive networks.</p
Higher-order reified adaptive network models with a strange loop
Full text linkIn this chapter, as in Chap. 7, the challenge of exploring plausible reified network models of order higher than two is addressed. This time another less usual option for application was addressed: the notion of Strange Loop which from a philosophical perspective sometimes is claimed to be at the basis of human intelligence and consciousness. This notion will be illustrated by examples from music, graphic art and paradoxes, and by Hofstadter’s claims about how Strange Loops apply to the brain. A reified adaptive network model of order higher than 2 was found, that even can be considered as being of infinite order. An example simulation shows the upward and downward interactions between the different levels, together with the processes within the levels. Another example addresses adaptive decision making according to two levels that are mutually reifying each other, as in Escher’s Drawing Hands lithograph.</p
- …
