Panagis Magdalinos

Enhancing a Fuzzy Logic Inference Engine through Machine Learning for a Self- Managed Network

Existing network management systems have static and predefined rules or parameters, while human intervention is usually required for their update. However, an autonomic network management system that operates in a volatile network environment should be able to adapt continuously its decision making mechanism through learning from the system’s behavior. In this paper, a novel learning scheme based on the network wide collected experience is proposed targeting the enhancement of network elements’ decision making engine. The algorithm employs a fuzzy logic inference engine in order to enable self-managed network elements to identify faults or optimization opportunities. The fuzzy logic engine is periodically updated through the use of two well known data mining techniques, namely k-Means and k-Nearest Neighbor. The proposed algorithm is evaluated in the context of a load identification problem. The acquired results prove that the proposed learning mechanism improves the deduction capability, thus promoting our algorithm as an attractive approach for enhancing the autonomic capabilities of network elements.

Tutamen: An Integrated Personal Mobile and Adaptable Video Platform for Health and Protection

A framework for mobile and portable High-Definition Video streaming is proposed, developed and assessed. Suitable for emergency scenarios, involving for instance ambulances and fire-fighters, the presented framework resorts to a state-of-art platform which considers off-the-shelf hardware and available video codecs for High-Definition Video. The obtained results show that the proposed architecture is able to efficiently support rescuing teams in the demanding scenarios where they operate, guaranteeing video quality and ease of use. This solution is particularly useful for situations where experts in the fields can accurately provide their insights and contributions remotely and in a timely fashion.

A Sequential Sampling Framework for Spectral k-Means Based on Efficient Bootstrap Accuracy Estimations: Application to Distributed Clustering

The scalability of learning algorithms has always been a central concern for data mining researchers, and nowadays, with the rapid increase in data storage capacities and availability, its importance has increased. To this end, sampling has been studied by several researchers in an effort to derive sufficiently accurate models using only small data fractions. In this article we focus on spectral k-means, that is, the k-means approximation as derived by the spectral relaxation, and propose a sequential sampling framework that iteratively enlarges the sample size until the k-means results (objective function and cluster structure) become indistinguishable from the asymptotic (infinite-data) output. In the proposed framework we adopt a commonly applied principle in data mining research that considers the use of minimal assumptions concerning the data generating distribution. This restriction imposes several challenges, mainly related to the efficiency of the sequential sampling procedure. These challenges are addressed using elements of matrix perturbation theory and statistics. Moreover, although the main focus is on spectral k-means, we also demonstrate that the proposed framework can be generalized to handle spectral clustering. The proposed sequential sampling framework is consecutively employed for addressing the distributed clustering problem, where the task is to construct a global model for data that resides in distributed network nodes. The main challenge in this context is related to the bandwidth constraints that are commonly imposed, thus requiring that the distributed clustering algorithm consumes a minimal amount of network load. This illustrates the applicability of the proposed approach, as it enables the determination of a minimal sample size that can be used for constructing an accurate clustering model that entails the distributional characteristics of the data. As opposed to the relevant distributed k-means approaches, our framework takes into account the fact that the choice of the number of clusters has a crucial effect on the required amount of communication. More precisely, the proposed algorithm is able to derive a statistical estimation of the required relative sizes for all possible values of k. This unique feature of our distributed clustering framework enables a network administrator to choose an economic solution that identifies the crude cluster structure of a dataset and not devote excessive network resources for identifying all the “correct” detailed clusters.

SYSTAS: Density-based algorithm for clusters discovery in wireless networks

The Internet of Things will comprise billions of randomly placed devices, forming a dense and unstructured network environment with overlapping wireless topologies. In such demanding environment, the grouping of IoT devices into clusters is a promising approach for the management and the control of network resources in the context of an autonomous system. This paper proposes the SYSTAS algorithm for the distributed discovery and formation of clusters in random geometric graphs of fixed wireless nodes by exploiting local topology knowledge and without having any information about the expected number of clusters. The density of the network graph, discovered by interacting with neighboring nodes and the topological features, as well as the model of preferential attachment are used by the proposed scheme. The effectiveness of SYSTAS is evaluated in various topologies. Experimental evaluation demonstrates that SYSTAS outperforms other clustering schemes; in some occasions these solutions have comparable results with SYSTAS but they require global network view, which leads to higher signaling cost.

Adapting policy-based management of Future Networks using collaborative filtering techniques

Future Networks constitute a complex and dynamic environment which network operators are called to orchestrate uniformly and efficiently. Among others, the increase in the number and heterogeneity of network infrastructure, the complexity of devices and protocols and the explosion in traffic demands are only “few” of the issues that network operators should take into consideration. Conventional network management schemes lack in automation, harmonization and efficiency, in order to handle such chaotic environments. Autonomic network management targets the governance of the behavior of autonomic and contemporary network entities, based on network operator requirements and business goals. Policies are considered as an effective tool for accomplishing a desirable high level control. In this paper, we present a novel policy-based network management framework, while enriching its ontology-oriented inference engine with collaborative filtering capabilities thus achieving the acceleration of the decision making process.

Coverage and Capacity Optimization of Self-Managed Future Internet Wireless Networks

Future Internet network management systems are expected to incorporate self-x capabilities in order to tackle the increased management needs that cannot be addressed through human intervention. Towards this end, Self-NET developed a self-management framework based on the introduction of cognitive capabilities in network elements. In this paper, the experimentation platform for “Coverage and Capacity Optimization of Self-managed Future Internet Wireless Network”, incorporating the self-management framework of Self-NET, is presented.

A context extraction and profiling engine for 5G network resource mapping

Abstract Future 5G network ecosystems comprise a plethora of 3GPP and non 3GGP Radio Access Technologies - RATs. Deployment scenarios envision a multi-layer use of macro, micro and femto-cells where multi-mode end devices, supporting different applications, are served by different technologies. The association of end devices to the most appropriate RAT/layer will therefore become a tantalizing process necessitating the introduction of mechanisms that decide and execute an optimal mapping. The latter is of paramount importance since sub-optimal configuration of network components will affect overall network performance. Towards this end, we introduce the Context Extraction and Profiling Engine (CEPE), a knowledge discovery (KDD) framework catering for the extraction and exploitation of user behavioral patterns from network and service information. An eNB exploits the knowledge scheme derived by CEPE in order to improve the placement of end devices to RATs/layers. In the context of this paper, we provide a thorough analysis of existing standards, research papers and patents, discuss the main innovation of our proposal and highlight the differences with existing schemes. Building on use cases involving mobility management mechanisms that typically affect device to technology mapping (i.e. cell (re)selection, handover) we provide an extensive set of experiments that demonstrate the validity and viability of our idea. Overall evaluation showcases that CEPE achieves high quality results thus emerging as a viable approach for network optimization in future 5G environments.

Context Monitoring Optimization in Autonomic Networks

The recent advances in network management systems suggest the adoption of autonomic mechanisms in order to minimize the need for human intervention while handling complex heterogeneous networks. Data acquisition performed by monitoring processes is an essential part of autonomic mechanisms. The rate of sampling is a crucial factor since it is related to (1) the successful/unsuccessful detection of events, (2) the processing power needed to perform the sampling and (3) the energy that a node consumes during such actions. In order to address these issues we designed a simple and efficient mechanism that dynamically adapts the sampling rate of the context monitoring procedure. The merits of the mechanism are quantified by means of an analytical model as well as through extensive simulations that validated the theoretic outcomes. Finally, we experimentally assessed the effectiveness and efficiency of our approach through two real-world experiments. Overall results showcase that our mechanism achieves high detection rates while in parallel minimizes significantly the number of monitoring loops thus, emerges as a viable approach for context monitoring optimization in autonomic networks.

Integrating the self-growing concept in a self-organizing wireless network for topology optimization

SUMMARY The concept of self-organizing networks is considered one of the most promising approaches for the efficient management of future wireless networks that will support a large number of nodes and a plethora of services with diverse characteristics. Today, different types of networks (e.g. WLANs, wireless sensor networks) are deployed to serve different needs but do not interoperate. Their possible loose integration will provide opportunities that could be exploited through collaborative approaches to devise novel solutions to extend the capabilities and improve the performance of these networks. The self-growing paradigm addresses this challenge by extending network nodes to dynamically evolve in terms of purpose and operational features. In this paper we describe the CONSERN architecture, which targets the realization of the self-growing concept in the context of self-organized networks. To test our ideas we designed and implemented a WLAN topology optimization scheme that provides the best coverage at a minimum energy consumption, through dynamic access point (AP) deactivation and reactivation. Using self-growing mechanisms and typical motion detectors we present how the operation of the proposed topology optimization mechanism can be improved. The reduced energy consumption attained under the proposed scheme at the AP side, as well as the efficient utilization of network resources, are evaluated via a proof-of-concept implementation that we have deployed in a real office environment that consists of WLAN APs and motion sensors. Copyright

A proof-of-concept testbed for cooperative and self-growing, energy aware networks

Cooperating heterogeneous wireless elements provide advanced problem solving capabilities and improved services. At the same time, low energy solutions create attractive business case offering significant benefits in terms of products dependability, and operation costs. CONSERN project addresses those challenges through the Self-growing paradigm, combining autonomic and collaborative capacities towards serving energy efficiency in heterogeneous networking environments. This paper describes the CONSERN Self-growing Functional Architecture, realised by a set of mechanisms and enablers which have been deployed in a real life proof-of-concept testbed.