Massimo Panella

Energy-efficient dynamic traffic offloading and reconfiguration of networked data centers for big data stream mobile computing: review, challenges, and a case study

Big data stream mobile computing is proposed as a paradigm that relies on the convergence of broadband Internet mobile networking and real-time mobile cloud computing. It aims at fostering the rise of novel self-configuring integrated computing-communication platforms for enabling in real time the offloading and processing of big data streams acquired by resource-limited mobile/wireless devices. This position article formalizes this paradigm, discusses its most significant application opportunities, and outlines the major challenges in performing real-time energy-efficient management of the distributed resources available at both mobile devices and Internet-connected data centers. The performance analysis of a small-scale prototype is also included in order to provide insight into the energy vs. performance tradeoff that is achievable through the optimized design of the resource management modules. Performance comparisons with some state-of-the-art resource managers corroborate the discussion. Hints for future research directions conclude the article.

Adaptive resolution min-max classifiers

A high automation degree is one of the most important features of data driven modeling tools and it should be taken into consideration in classification systems design. In this regard, constructive training algorithms are essential to improve the automation degree of a modeling system. Among neuro-fuzzy classifiers, Simpsons (1992) min-max networks have the advantage of being trained in a constructive way. The use of the hyperbox, as a frame on which different membership functions can be tailored, makes the min-max model a flexible tool. However, the original training algorithm evidences some serious drawbacks, together with a low automation degree. In order to overcome these inconveniences, in this paper two new learning algorithms for fuzzy min-max neural classifiers are proposed: the adaptive resolution classifier (ARC) and its pruning version (PARC). ARC/PARC generates a regularized min-max network by a succession of hyperbox cuts. The generalization capability of ARC/PARC technique mostly depends on the adopted cutting strategy. By using a recursive cutting procedure (R-ARC and R-PARC) it is possible to obtain better results. ARC, PARC, R-ARC, and R-PARC are characterized by a high automation degree and allow to achieve networks with a remarkable generalization capability. Their performances are evaluated through a set of toy problems and real data benchmarks. The paper also proposes a suitable index that can be used for the sensitivity analysis of the classification systems under consideration.

Distributed learning for Random Vector Functional-Link networks

This paper aims to develop distributed learning algorithms for Random Vector Functional-Link (RVFL) networks, where training data is distributed under a decentralized information structure. Two algorithms are proposed by using Decentralized Average Consensus (DAC) and Alternating Direction Method of Multipliers (ADMM) strategies, respectively. These algorithms work in a fully distributed fashion and have no requirement on coordination from a central agent during the learning process. For distributed learning, the goal is to build a common learner model which optimizes the system performance over the whole set of local data. In this work, it is assumed that all stations know the initial weights of the input layer, the output weights of local RVFL networks can be shared through communication channels among neighboring nodes only, and local datasets are blocked strictly. The proposed learning algorithms are evaluated over five benchmark datasets. Experimental results with comparisons show that the DAC-based learning algorithm performs favorably in terms of effectiveness, efficiency and computational complexity, followed by the ADMM-based learning algorithm with promising accuracy but higher computational burden.

An input-output clustering approach to the synthesis of ANFIS networks

A useful neural network paradigm for the solution of function approximation problems is represented by adaptive neuro-fuzzy inference systems (ANFIS). Data driven procedures for the synthesis of ANFIS networks are typically based on clustering a training set of numerical samples of the unknown function to be approximated. Some serious drawbacks often affect the clustering algorithms adopted in this context, according to the particular data space where they are applied. To overcome such problems, we propose a new ANFIS synthesis procedure where clustering is applied in the joint input-output data space. Using this approach, it is possible to determine the consequent part of Sugeno first-order rules and therefore the hyperplanes characterizing the local structure of the function to be approximated. Successively, the fuzzy antecedent part of each rule is determined using a particular fuzzy min-max classifier, which is based on the adaptive resolution mechanism. The generalization capability of the resulting ANFIS architecture is optimized using a constructive procedure for the automatic determination of the optimal number of rules. Simulation tests and comparisons with respect to other neuro-fuzzy techniques are discussed in the paper, in order to assess the efficiency of the proposed approach.

An RNS Architecture for Quasi-Chaotic Oscillators

Wideband chaotic carrier is a promising solution for wideband communication, since it overcomes the disadvantages of both narrowband and spread-spectrum communication. It is particularly suited to realize information encryption for secure communication. Chaotic signals can be generated by using discrete-time non-linear dynamical circuits, since they can exhibit a quasi-chaotic (QC) behavior. A particular implementation of QC digital filters can be based on finite precision arithmetic and, in particular, on residue number system (RNS) circuits, which possess very attractive features with regard to their VLSI implementation. In the present paper, we propose an RNS architecture that can be used in connection with secure communication. Each RNS channel consists of a QC oscillator, having its coefficients belonging to a Galois field defined by a prime modulus. In particular, the QC behavior is ensured by well-known properties of primitive polynomials in this field, which generate the characteristic feedback of the oscillator. We demonstrate in the paper that the proposed RNS architecture yields a cost-effective VLSI implementation, which favorably compares with respect to other secure communication approaches proposed in the technical literature. We obtain encouraging results both in terms of confidentiality of the encrypted information and of throughput rate for real-time applications. Moreover, we propose an extended architecture suited to the protection of the secure communication system against transmission errors, by using the self-correcting ability of Redundant RNS (RRNS).

A decentralized training algorithm for Echo State Networks in distributed big data applications

The current big data deluge requires innovative solutions for performing efficient inference on large, heterogeneous amounts of information. Apart from the known challenges deriving from high volume and velocity, real-world big data applications may impose additional technological constraints, including the need for a fully decentralized training architecture. While several alternatives exist for training feed-forward neural networks in such a distributed setting, less attention has been devoted to the case of decentralized training of recurrent neural networks (RNNs). In this paper, we propose such an algorithm for a class of RNNs known as Echo State Networks. The algorithm is based on the well-known Alternating Direction Method of Multipliers optimization procedure. It is formulated only in terms of local exchanges between neighboring agents, without reliance on a coordinating node. Additionally, it does not require the communication of training patterns, which is a crucial component in realistic big data implementations. Experimental results on large scale artificial datasets show that it compares favorably with a fully centralized implementation, in terms of speed, efficiency and generalization accuracy.

Partial discharge pattern recognition by neuro-fuzzy networks in heat-shrinkable joints and terminations of XLPE insulated distribution cables

An identification technique is described, based on a developed adaptive fuzzy logic network, that enables the recognition of partial discharges (PD) generated by different defects in heat-shrinkable joints and terminations of XLPE insulated power distribution cables. It is shown that different sources of PD can be identified on the basis of fuzzy rules applied to a selection of parameters derived from PD-pulse phase and amplitude distributions. A comparison with other PD pattern recognition techniques based on traditional neural networks is presented and discussed.

Forecasting Energy Commodity Prices Using Neural Networks

A new machine learning approach for price modeling is proposed. The use of neural networks as an advanced signal processing tool may be successfully used to model and forecast energy commodity prices, such as crude oil, coal, natural gas, and electricity prices. Energy commodities have shown explosive growth in the last decade. They have become a new asset class used also for investment purposes. This creates a huge demand for better modeling as what occurred in the stock markets in the 1970s. Their price behavior presents unique features causing complex dynamics whose prediction is regarded as a challenging task. The use of a Mixture of Gaussian neural network may provide significant improvements with respect to other well-known models. We propose a computationally efficient learning of this neural network using the maximum likelihood estimation approach to calibrate the parameters. The optimal model is identified using a hierarchical constructive procedure that progressively increases the model complexity. Extensive computer simulations validate the proposed approach and provide an accurate description of commodities prices dynamics.

Neural networks with quantum architecture and quantum learning

A method is proposed for solving the two key problems facing quantum neural networks: introduction of nonlinearity in the neuron operation and efficient use of quantum superposition in the learning algorithm. The former is indirectly solved by using suitable Boolean functions. The latter is based on the use of a suitable nonlinear quantum circuit. The resulting learning procedure does not apply any optimization method. The optimal neural network is obtained by applying an exhaustive search among all the possible solutions. The exhaustive search is carried out by the proposed quantum circuit composed of both linear and nonlinear components. Copyright

Refining accuracy of environmental data prediction by MoG neural networks

Abstract The prediction of future values of environmental data sequences is mandatory to the cost-effective management of available resources. Consequently, the possibility to improve the prediction accuracy is a very important goal to be pursued. We propose in the present paper two possible approaches for refining the prediction accuracy on real data sequences. Both these approaches make use of Mixture of Gaussian neural networks for the solution of suitable function approximation problems. The first approach pursues the regularization of the learning process based on the reconstructed state of the context delivering the sequence; the second one is based on the particular chaotic nature of the prediction error.