Mechanisms on automatic discovery of macro actions or skills in reinforcement learning methods are mainly focused on subgoal discovery methods. Among the proposed algorithms, those based on graph centrality measures demonstrate a high performance gai

Many algorithms have recently been proposed for finding communities in networks. By definition, a community is a subset of vertices with a high number of connections among the vertices, but only few connections with other vertices. The worst drawback of most of the proposed algorithms is their computational complexity which is usually an exponentially increasing function of the number of the vertices. Newman-Fast is a well-known community detection algorithm which is suitable for large networks due to its low computational cost. Although the performance of this algorithm is good for well structured networks, it does not perform well for more fuzzy-clustered networks. In this paper, we propose a weighting scheme which considerably enhances the performance of the Newman-Fast algorithm with a little effort. We also show that the modified algorithm effectively enhances the community discovery process in both computer-generated and real-world networks.

A community in a network is a subset of vertices densely connected to each other, but less connected to the vertices outside. Many different approaches have been developed to find such structures in a given network, but the main drawback of most of the available algorithms is that they are computationally demanding and their complexity is usually an exponentially increasing function of the number of the vertices. Newman-Fast is a well-known community detection algorithm which is suitable for large networks due to its low computational cost. Although the performance of this algorithm is good for well structured networks, it does not perform well for more fuzzy-clustered networks. In this paper, we utilize a proper weighting scheme and an algorithm based on partial synchronization phenomenon as pre- and post-processing steps to improve the Newman-Fast algorithm. Furthermore, we evaluate the proposed method for both computer-generated and real-world networks. The results show that either both or one of the proposed steps enhance the performance of the Newman-Fast algorithm significantly while they impose little additional effort.

Reinforcement Learning (RL) is an approach for training agent's behavior through trial-and-error interactions with a dynamic environment. An important problem of RL is that in large domains an enormous number of decisions are to be made. Hence, instead of learning using individual primitive actions, an agent could learn much faster if it could form high level behaviors known as skills. Graph-based approach, that maps the RL problem to a graph, is one of the several approaches proposed to identify the skills to learn automatically. In this paper we propose a new centrality measure for identifying bottleneck nodes crucial to develop useful skills. We will show through simulations for two benchmark tasks, namely, “two-room grid” and “taxi driver” that a procedure based on the proposed measure performs better than the procedure based on closeness and node betweenness centrality.

To characterize the behavior and robustness of cellular circuits with many unknown parameters is a major challenge for systems biology. Its difficulty rises exponentially with the number of circuit components. We here propose a novel analysis method to meet this challenge. Our method identifies the region of a high-dimensional parameter space where a circuit displays an experimentally observed behavior. It does so via a Monte Carlo approach guided by principal component analysis, in order to allow efficient sampling of this space. This 'global' analysis is then supplemented by a 'local' analysis, in which circuit robustness is determined for each of the thousands of parameter sets sampled in the global analysis. We apply this method to two prominent, recent models of the cyanobacterial circadian oscillator, an autocatalytic model, and a model centered on consecutive phosphorylation at two sites of the KaiC protein, a key circadian regulator. For these models, we find that the two-sites architecture is much more robust than the autocatalytic one, both globally and locally, based on five different quantifiers of robustness, including robustness to parameter perturbations and to molecular noise. Our 'glocal' combination of global and local analyses can also identify key causes of high or low robustness. In doing so, our approach helps to unravel the architectural origin of robust circuit behavior. Complementarily, identifying fragile aspects of system behavior can aid in designing perturbation experiments that may discriminate between competing mechanisms and different parameter sets.

In this paper, we explain how to build a turbo-like structure with binary inputs and chaotic outputs for efficient coding and decoding in additive white Gaussian noise (AWGN). We analyze the convergence of the decoding algorithm, the performance in the error floor region and explain minimum distance properties of the resulting codes.

Mixed-signal systems combine analog and digital signal processing in one overall design that should be integrated as a single system on chip. At system level, the separation into isolated subsystem representations for analog and digital parts needs to be overcome by a unified formalism that can handle any mix of continuous-time, discrete-time, and hybrid systems. A second driving force in our contribution is that, with decreasing circuit size, nonlinear effects become increasingly prominent. Therefore, several aspects of a consistent theory of nonlinear dynamical systems with mixed-signal domains have been developed. It is based on an extension of the Volterra series approach to the representation of fading-memory systems. We will present novel mixed-domain representations, both in the time and frequency domains, and we develop simple results for system composition. For illustration purposes, we discuss analog-to-digital conversion, digital-to-analog conversion, and the digital cancellation of a (nonlinear) analog echo path.

In this paper we investigate optimization of synchronization cost in undirected dynamical networks. To do so, proper weights are assigned to the networks' edges considering node and edge betweenness centrality measures. The proposed method gives near-optimal results with less complexity of computation compared to the optimal method, i.e. the method based on convex optimization. This property enables us to apply the method to large networks. Through numerical simulations on scale-free and small-world networks of different size and topological properties we give evidence that the performance of the proposed method is much better than another heuristic method; namely the Metropolis-Hasting algorithm. This procedure has potential application in many engineering problems where the synchronization of the network is required to be achieved by minimal cost.

In this paper, using the mathematical properties of self-kronecker-production of small size random matrices, a simple but effective method is presented to optimize the reservoir of an echo state network given a certain task. The experimental results investigating the NARMA system show that few steps of the proposed optimization process can lead to a near optimum solution.

It is investigated to what extent the trajectories of a stochastically switched (blinking) system follow the corresponding trajectories of the averaged system. Four cases have to be distinguished, depending on whether or not the averaged system has a unique attractor and whether or not the attractor(s) is (are) invariant under the dynamics of the blinking system. The corresponding asymptotic behavior of the trajectories of the blinking system is described and illustrative examples are given

In this paper a class of discretized piecewise linear chaotic maps of a very high dimensions are used for communication over a noisy channel. An information payload that is sent over a channel is controlled in the transmitter and is related to the symbolic sequences of chaotic trajectory. Such chaotic systems are used for generating new ensembles of nonlinear codes. The proposed decoding method is based on belief propagation principle, such that a complete link to modern coding theory is established. The transmission of information using chaotic systems, usually showing mediocre performance, here on contrary, have a very good performance with a threshold for a certain level of signal-to-noise ratio (SNR)

We do a case study of two different analysis techniques for studying the stochastic behavior of a randomized system/algorithms: (i) The first approach can be broadly termed as a mean value analysis (MVA), where the evolution of the mean state is studied assuming that the system always actually resides in the mean state; (ii) The second approach looks at the probability distribution function of the system states at any time instance, thus studying the evolution of the (probability mass) distribution function (EoDF).

We consider a large-scale wireless network, but with a low density of nodes per unit area. Interferences are then less critical, contrary to connectivity. This paper studies the latter property for both a purely ad-hoc network and a hybrid network, where fixed base stations can be reached in multiple hops. We assume here that power constraints are modeled by a maximal distance above which two nodes are not (directly) connected. We find that the introduction of a sparse network of base stations does significantly help in increasing the connectivity, but only when the node density is much larger in one dimension than in the other. We explain the results by percolation theory. We obtain analytical expressions of the probability of connectivity in the 1D case. We also show that at a low spatial density of nodes, bottlenecks are unavoidable. Results obtained on actual population data confirm our findings.

----

The paper describes the state of the art in the design of receivers for chaos-based digital communication over noisy channels. An information theoretic analysis of the potential of chaos in digital communication schemes is given, underlining that there is no fundamental principle that speaks against the use of chaos in digital communications. The design of the optimal receiver based on statistically optimal detection is shown. The channel model considered throughout the paper is that of additive white Gaussian noise. While the optimal receiver provides the base for assessing the potential of chaos-based schemes in this application, suboptimal versions thereof as presented in the paper allow efficient implementations.

The inverse system approach is an uniform view on hiding and retrieving information from chaotic signals. A clue to the understanding of the system inversion is the relative degree and in connection with it a state transformation into a normal form, both presented in the paper. Inverse system examples published so far are classified with respect to this uniform view. A general structure for system inversion is introduced and applied in a novel circuit example.

The various special features of cellular neural networks are presented. It is stated what has been achieved so far and what future work is still needed in order to obtain a truly universal building block for information processing systems.>