Modeling Complex Systems by Structural Invariants Approach

When modeling complex systems, we usually encounter the following difficulties: partiality, large amount of data, and uncertainty of conclusions. It can be said that none of the known approaches solves these difficulties perfectly, especially in cases where we expect emergences in the complex system. The most common is the physical approach, sometimes reinforced by statistical procedures. The physical approach to modeling leads to a complicated description of phenomena associated with a relatively simple geometry. If we assume emergences in the complex system, the physical approach is not appropriate at all. In this article, we apply the approach of structural invariants, which has the opposite properties: a simple description of phenomena associated with a more complicated geometry (in our case pregeometry). It does not require as much data and the calculations are simple. The price paid for the apparent simplicity is a qualitative interpretation of the results, which carries a special type of uncertainty. Attention is mainly focused (in this article) on the invariant matroid and bases of matroid (M, BM) in combination with the Ramsey graph theory. In addition, this article introduces a calculus that describes the emergent phenomenon using two quantities—the power of the emergent phenomenon and the complexity of the structure that is associated with this phenomenon. The developed method is used in the paper for modeling and detecting emergent situations in cases of water floods, traffic jams, and phase transition in chemistry.

Structural Invariants in XForms

XForms is a declarative XML-based programming language for writing applications for the web and elsewhere. One of the central aspects of XForms is invariants that describe relationships between values, such that if a value changes or is changed, its related values specified in the invariants get updated automatically. This is much like how spreadsheets work, though more general. A major advantage of this approach is that much administrative detail is taken out of the hands of the programmer, and done automatically: the programmer specifies the relationships, and the computer does the work. However, XForms in its current incarnation only allows invariants to be placed between simple content values, even though there are important relationships that could be specified over data structures as a whole. This paper explores the possibilities for extending the mechanism to more general cases.

Phase Transition as an Emergent Phenomenon Analysed by Violation of Structural Invariant (M, BM)

When modeling complex systems, we usually encounter the following difficulties: partiality, large amounts of data and uncertainty of conclusions. The most common approach used for modeling is the physical approach, sometimes reinforced by statistical procedures. If we assume emergences in the complex system, a physical approach is not appropriate at all. Instead, we build here the approach of structural invariants. In this paper, we show that another plane can be built above the plane of physical description, which is responsible for violation of structural invariants. Main attention is concentrated (in this article) on the invariant matroid and bases of matroid (M, BM) in combination with Ramsey graph theory. In addition, the article introduces a calculus that describes the emergent phenomena using two quantities - the power of the emergent phenomenon and the complexity of the structure of the considered complex system. We show the application of the method for modeling phase transition in chemistry.

Structure-preserving discretization and control of a two-dimensional vibro-acoustic tube

This paper deals with the structure-preserving discretization and control of a two-dimensional vibro-acoustic tube using the port-Hamiltonian framework. A discretization scheme is proposed, and a set of precise basis functions are given in order to obtain a structure-preserving finite-dimensional port- Hamiltonian approximation of the two-dimensional vibro-acoustic system. Using the closed-loop structural invariants of the approximated system an energy-Casimir controller is derived. The performance of the proposed discretization scheme and the controller is shown by means of numerical simulations.

Self-Attentional Credit Assignment for Transfer in Reinforcement Learning

The ability to transfer knowledge to novel environments and tasks is a sensible desiderata for general learning agents. Despite the apparent promises, transfer in RL is still an open and little exploited research area. In this paper, we take a brand-new perspective about transfer: we suggest that the ability to assign credit unveils structural invariants in the tasks that can be transferred to make RL more sample-efficient. Our main contribution is SECRET, a novel approach to transfer learning for RL that uses a backward-view credit assignment mechanism based on a self-attentive architecture. Two aspects are key to its generality: it learns to assign credit as a separate offline supervised process and exclusively modifies the reward function. Consequently, it can be supplemented by transfer methods that do not modify the reward function and it can be plugged on top of any RL algorithm.