Training neural networks has become an almost daily activity for researchers working on different fields. Preparing accurately the training patterns often appears to be fundamental in obtaining a good balance between learning performance and time needed to reach them. In this contribution, we explore a paradigm to organize patterns for training quaternion neural networks for image processing tasks. The basic idea is to exploit the working principle of Cellular Nonlinear Networks, where local interactions are fundamental, to determine an efficient learning of multidimensional neural network, thus merging the main characteristics of the two architectures. A robustness analysis and practical applications are also presented.

Due to their synaptic-like characteristics and memory properties, memristors are often used in neuromorphic circuits, particularly neural network circuits. However, most of the existing neural network circuits that can generate complex dynamics have high dimensions and excessive connections, which is not conducive to implementation. This article introduces a memristor containing an arctangent function into a simple cyclic neural network (SCNN) circuit to design a simple cyclic memristive neural network (SCMNN) circuit capable of generating complex multiscroll chaotic attractors. The designed SCMNN contains an external stimulus current and generates multiscroll attractors, with the number of scrolls expanding as the switches in the memristor equivalent circuit are activated. By varying the parameters, the multiscroll attractors can be broken into different numbers of coexisting attractors, which also depends on the switch, and it can achieve multivariable amplitude control when there is only one scroll. The anti-interference ability of the circuit is tested. A low-cost circuit-based microcontroller suitable for engineering applications is designed for it, and multiscroll attractors are successfully captured in an oscilloscope. The National Institute of Standards and Technology (NIST) test is carried out to verify its application value.

In view of the superiority of memristors in strengthening dynamical complexity and the significant application potiential of multiscroll chaos, this paper attempts to introduce two memristors with scalable memductances into simple seed chaotic system for designing multiscroll memristive chaotic system (MMCS). The designed MMCS yields hidden grid multiscroll chaotic attractors with any number of scrolls expanding along with the internal variables of memristors. By varying the parameters, the multiscroll attractors can be broken into coexisting attractors with different numbers and scrolls dependent on parameters, and their oscillation amplitudes can be increased (or decreased) without changing the chaotic features. Dynamical analysis and circuit implementation are given to reveal the complexity and feasibility of the MMCS. The fixed-time synchronization (FxTS) is studied by using adaptive controller and the sufficient condition for FxTS is established via Lyapunov stability theory (LST). A multilevel secure communication scheme based on the FxTS of MMCS is designed and the experimental tests on the image, audio and data secure communication verify its effectiveness, which to some extent shows the application availability of MMCS.

This contribution is oriented towards proposing a novel strategy to simulate and predict planetary orbits based on exploiting the optimized performance of hypercomplex neural networks. These structures perform a learning process based on quaternion algebra thus leading to lower the number of epochs required to obtain an adequate degree of approximation. Moreover, we propose a strategy for predicting the outcome of the learning phase by instantiating suitable hypercomplex auxiliary networks to predict the trends of the main weights of the network.

Neural networks based on back-propagation learning algorithms and gradient descent algorithms are the first and the easiest tools developed for machine learning. They are still widespread nowadays, so much so by exploiting a huge number of different coding languages, between which MatLab, Python or Java, we have the possibility of using these training tools. But as highlighted in the past, these traditional neural networks suffer from their slow convergence rate. Aim of this paper is to revisit an algorithm to improve the speed of the learning phase, by exploiting the power of parallel computing to train a suitable number of auxiliary neural networks which work concurrently with the principal network. The implementation of the proposed algorithm in MatLab is shown in order to make evident the main difference with the traditional learning algorithms. Several examples, related to modeling of technological datasets from industrial environment, confirm the suitability of the proposed procedure.

Model-based approaches aiming to characterize human behavior when interacting with a controlled machine have been a matter of research investigation in various domains, from aerospace to semi-autonomous driving and robotics. Human-robot collaboration is one of the most exciting scenarios of application in which a continuous physical interaction between humans and the controlled plant is present. In this context, the human subject can adapt its control behavior to the external sensed dynamics. This capability has a significant observable effect on the control delay, making its characterization and prevision a crucial aspect to understand. This work investigates a linear modeling approach that uniquely describes human and robot control actions and applies to a collaborative robotic task.

This paper presents a Model-Based Design (MBD) methodology as a promising approach for the rapid and se-cure development of Embedded System applications, including those involving digital controllers. Based on existing hardware and software application-oriented tools by STMicroelectronics, a new modeling technique has been implemented to move from a traditional design workflow to an MBD one by using Mathworks®software platform. For a practical application of the proposed approach, a case study is reported to design a Permanent Magnet Synchronous Motor (PMSM) drive based on the STM32 MCU family. The main contribution of this paper is to make focus on concretely building a unique model architecture of a complex software/hardware system able to implement all the major aspects of MBD methodology such as executable requirements, Normal and Processor-In-the-Loop (PIL) simulation modes, continuous test and verification, automatic code generation. To achieve this purpose, a new Simulink®blockset dedicated to the STM32 Motor Control ecosystem was developed. It includes Simulink blocks for maths, algorithms, and IPs, electronic circuitries, MCU peripherals, speed sensors. Simulink®Embedded Coder tool has been used to automatically generate code for a specific STM32 MCU tar-get to overcome the time consuming and error-prone problems of the handwritten coding.

Plasma physics is an example of research field where many measurements carried out at very specific working conditions need to be collected and processed. By looking at the properties of these data, it can be possible to explore their hidden features in order to solve challenging problems that usually require high computational efforts, such as the tomographic reconstruction. In this paper, preliminary but nontrivial analyses of flux measurements produced in a Tokamak machine are shown and discussed, with the aim of introducing an application of some algorithms for feature selection to detect hidden, relevant relationships within given sets of channels. All the statistical details, and therefore the feature selection procedure itself, are introduced in view of further deepenings, such as the aforementioned problem of tomographically reconstructing plasma profiles from flux measurements or modelling the system in terms of its input-output relationship.

The concept of imperfection is always associated to a negative outlook. However, real devices are always imperfect and operate far from ideality. Despite this, real devices actually work. This is due to the fact that imperfections give rise to hidden dynamics, which can be excited to gain a positive effect on the overall behavior of the device. Human-machine systems are relevant examples of imperfect systems in which the presence of imperfections plays a crucial role, either negative or positive. In this paper, we focus on a paradigmatic example of imperfect systems, represented by a complex and imperfect electromechanical structure which supports and couples 15 coils rotating thanks to the electromagnetic interaction with associated magnets. The electrical and mechanical interaction between the coils and that between the coils and the structure generate complex patterns of vibration which may prevent the system from reaching the correct working conditions, i.e., all coils actually rotating. A control strategy to ensure coils rotation based on the excitation of the hidden dynamics induced by imperfections is discussed, characterizing its effect with respect to the control signal properties and to the power provided to the structure. It is worth to notice that the control strategy is essentially based on the excitation rather than on the suppression of the hidden dynamics, thus providing evidence on the possibility to control imperfect systems exploiting imperfections.

In this paper, a fractional order representation and the related hardware realization of the Duffing system are presented. The considered system can be defined as the general case of the standard Duffing system, which is a typical example of a non autonomous chaotic system. The aim of this paper is to confirm that an opportunely excited fractional order Duffing system, with order less than two, can have a chaotic behavior. This statement is supported with the circuital implementation of the circuit.

The past few decades have seen an explosion of satellite remote sensing techniques for the monitoring of volcanic thermal features. Here, we propose an artificial neural network approach for improving the cloud detection through imperfect multispectral satellite images analysis. The cloud detection algorithm has been tested on a data set of MSG-SEVIRI images acquired over the area of Etna volcano in Sicily (Italy) before and during the 2008 eruption. Results show that this approach is robust in terms of percentage of correctly classified pixels.

System-of-Systems (SoS) can be defined as a large-scale integration of many independent, self-contained systems having the common aim of satisfying a global need. Under this perspective, lots of systems of systems can be found in several fields, where a common final goal drives systems towards a final SoS state. In nuclear fusion research area, several examples of SoS applications can be made. From the integration of all the constituents of the TOKAMAK machines, which work together to achieve a sustained nuclear fusion reaction, to the circuits made of active analogue components mimicking plasma behavior, to the neural networks made of connected units working together in order to predict plasma variables behavior. In this paper, we aim at collecting several examples of how a SoS approach can be adopted in order to characterize and control TOKAMAK relevant scenarios.

In this paper we study the effects of noise in computing devices composed by coupled memristive oscillators. In particular, we will focus the attention on the emergence of spatiotemporal chaos in a computing architecture made of simple oscillators composed by three basic elements, namely a capacitor, an inductor and a memristor, driven by a sinusoidal external signal. The characterization of the dynamical behavior of the simple oscillator in terms of flux and charge will be discussed and the counterintuitive role of initial conditions on memristors will be highlighted. Aim of this contribution is to provide evidence that a robust synchronized emergent behavior can be attained as a direct positive consequence of the presence of unavoidable sources of noise on initial conditions.

In this communication, the first example of fractional order nonlinear system showing jump resonance is described, that is a system in which the order of the derivative is non-integer and whose frequency response is a multi-valued function in a given range of frequencies. Furthermore, a strategy to design fractional order systems showing jump resonance is presented along with the procedure to design and implement an analog circuit based on the approximation of the fractional order derivative. An extensive numerical analysis allows to asses that the phenomenon is robust to difference in the derivative order, enlightening the fact that a system with order lower than two is able to provide a jump resonance behavior.

The study on stochastic effects occurring at the nanoscales in VLSI memristive devices is a timely topic which is gaining a growing interest. Indeed, these effects are often linked to the occurrence of nonlinear phenomena, such as spatiotemporal chaos. In this paper, we present evidence that spatiotemporal chaos in VLSI memristive systems can also be tamed by exploiting the unavoidable sources of noise affecting the dynamical behavior of the device. In particular, we aim at investigating this scenario starting from the results observed in several nonlinear oscillators, where the presence of noise facilitates their synchronization. In particular, we focus on the role of noise acting on initial conditions in memristive nonlinear oscillators. In this paper, we focus on a nonautonomous memristive oscillator characterizing its dynamics with respect to the memristor initial conditions and deriving a suitable model in which they appear as further system parameters and exploiting this dependence for the synchronization of an array of coupled nonlinear chaotic memristive oscillators.

In this paper, the model identification of the temperature over the surface of the limiter adopted in the Frascati Tokamak Upgrade (FTU) is presented. Tokamaks are considered as the most interesting facilities to study self-sustained nuclear fusion reactions. Recently, a Liquid Lithium Limiter (LLL) has been introduced in the FTU with the aim of reducing impurities in the plasma. However, the performance of the LLL are maximized when temperature over its surface is uniformly distributed. In this paper, we face the problem of modeling the thermal behavior of the limiter surface following two different data-driven approaches: a linear autoregressive model, and a nonlinear autoregressive model. A comparison among the two models will be given, showing also which physical quantities are relevant to the specific modeling problem.

The integration in small packages of high power modules requires in the phase of prototype testing integrated measurements and advanced software tools to assure high level reliability of the devices. In this paper a new integrated equipment based on the concept of system-of-systems is proposed. The system-of-systems equipment is intended as a platform including several measurement instrumentations, the power supply, the control boards and the testing modules, that themselves are complex systems.

In this paper, we introduce a network of spiking neurons devoted to navigation control. Three different examples, dealing with stimuli of increasing complexity, are investigated. In the first one, obstacle avoidance in a simulated robot is achieved through a network of spiking neurons. In the second example, a second layer is designed aiming to provide the robot with a target approaching system, making it able to move towards visual targets. Finally, a network of spiking neurons for navigation based on visual cues is introduced. In all cases, the robot was assumed to rely on some a priori known responses to low-level sensors (i.e., to contact sensors in the case of obstacles, to proximity target sensors in the case of visual targets, or to the visual target for navigation with visual cues). Based on their knowledge, the robot has to learn the response to high-level stimuli (i.e., range finder sensors or visual input). The biologically plausible paradigm of spike-timing-dependent plasticity (STDP) is included in the network to make the system able to learn high-level responses that guide navigation through a simple unstructured environment. The learning procedure is based on classical conditioning.

In this paper a recently introduced and well promising methodology for robot perception is applied to autonomously learn robot navigation in an unstructured environment. Perception is here considered as the spontaneous, environmentally mediated, emergence of Turing patterns in CNNs as perceptual states. They, plastically associated to suitable actions, lead the robot to solve autonomously its task. Following this concept, robot behavior (in this case navigation) is reflected, in a virtual navigation through the different basins of attraction of the generated patterns, within the robot control neural network. The whole architecture was implemented in an FPGA-based hardware embedded on a roving robot. In the manuscript the perceptual architecture together with experimental results on a roving robot, will be reported.

Many works investigated the phenomenon of lava flow through numerical models, obtaining excellent results, although most of the models require several approximations. Each model, in fact, has its restrictions: for instance, some of them work only on inclined planes, while others do not consider cooling processes associated to lava flow and so on. Simplifications are often needed to afford the computational effort required by the problem. Although the increasing computational capability of computers, due to technological progress, can be very useful to quickly resolve differential equations, which are essential to study lava flows, it is still important to take into account alternative solutions to the problem. One of these solutions is the use of parallel analog processors, namely Cellular Nonlinear Networks (CNNs). In particular, the approach used in this work is based on the E^3 architecture [1] used for the first time to study lava flows. Two different models are proposed.

Magneto encephalographic (MEG) brain signals are studied using a method for characterizing nonlinear dynamics. This approach uses the value of dinfin(d-infinite) to characterize the system's asymptotic chaotic behavior. A novel procedure was developed to extract this parameter from time series when the system's structure and laws are unknown. The implementation of the algorithm has proven to be general and computationally efficient. The information characterized by this parameter is furthermore independent and complementary to the signal power since it considers signals normalized with respect to their amplitude. The algorithm implemented here is applied to whole-head 148 channel MEG data during two highly structured yogic breathing meditation techniques. Results, which are relative to spatio-temporal distributions of the calculated dinfin on the MEG channels, are analyzed and compared during different phases of the yogic protocol.

Correlation based algorithms have been found to explain many basic behaviors in simple animals. In this paper the authors investigate the problem of navigation control of a robot from the viewpoint of bio-inspired perception. In this paper the authors study how to go up, through learning, from the implementation of a reactive system, towards behaviors of increasing complexity. The whole control system is based on networks of spiking neurons. A correlation based rule, namely the spike timing dependent plasticity (STDP), is implemented for an efficient learning. The main interesting consequence is that the system is able to learn high-level sensor features, based on a set of basic reflexes, depending on some low-level sensor inputs. The whole methodology is presented through simulation results and also through its implementation on an FPGA based system for real time working on a roving robot.

This paper describes a general approach for the adaptive supervised learning of behaviors in a behavior-based robot. The key idea is to formalize a behavior produced by a Motor Map driven by an internal adaptive reward function. Aim of the adaptive reward function is to select the most significant sensory inputs and to use them in the best way. The greatest challenge is to keep small the search space. Motor map learning relies on the classical Kohonen algorithm, while the structure of the reward function is learnt through a non-associative reinforcement learning algorithm. Simulation results on a six legged biologically-inspired robot confirm the suitability of the approach. This methodology allows the human designer to easily embody all the a priori knowledge on the robot controller, while providing at the same time a high degree of adaptability and robustness against the sensory malfunctioning

This paper describes the implementation of a bio-inspired six legged robot: Gregor I. Both structure and locomotion control are inspired by biological observations in cockroaches. Robot mechanics attempts to emulate main structural features in cockroaches, like self-stabilizing posture and specializing legged function; in turn, locomotion control is based on the theory of the central pattern generator implemented on a VLSI chip. The final aim is to artificially replicate the fundamental principles that guarantee cockroach's extraordinary agility. Our major concern was on the implementation of rear legs, that seem to play a crucial role in obstacle overcoming and payload capability, and on the locomotion control, performed in this work by a cellular neural network playing the role of an artificial central pattern generator. Experimental tests showed that Gregor I is able to walk at the travel speed of 0.1 body length per second and to successfully negotiate obstacles more than 170 % of the height of its mass center

This paper presents an innovative wormlike robot controlled by cellular neural networks (CNNs) and made of an ionic polymer-metal composite (IPMC) self-actuated skeleton. The IPMC actuators, from which it is made of, are new materials that behave similarly to biological muscles. The idea that inspired the work is the possibility of using IPMCs to design autonomous moving structures. CNNs have already demonstrated their powerfulness as new structures for bio-inspired locomotion generation and control. The control scheme for the proposed IPMC moving structure is based on CNNs. The wormlike robot is totally made of IPMCs, and each actuator has to carry its own weight. All the actuators are connected together without using any other additional part, thereby constituting the robot structure itself. Worm locomotion is performed by bending the actuators sequentially from "tail" to "head," imitating the traveling wave observed in real-world undulatory locomotion. The activation signals are generated by a CNN. In the authors' opinion, the proposed strategy represents a promising solution in the field of autonomous and light structures that are capable of reconfiguring and moving in line with spatial-temporal dynamics generated by CNNs.

In the last years, in the area of bio-inspired robotics, the research activity has been directed to high level aspects that includes psychological theories and behavioral approaches. In this paper a new simulation tool for perceptive system based on the sensing-perception-action loop is proposed. The framework has been designed to evaluate the performance of control strategies applied to the navigation of autonomous robots. The tool can be used to create a 3D environment in which the exploring capabilities of a robot executing a navigation task such as for example a food retrieval task, can be evaluated. The behavior of the system is monitored with the help of a 3D real-time visualizer supported by a graphic representations of the trajectory followed

In this work a nonlinear dynamical model of a fuel cell stack is developed by means of artificial neural networks. The model presented is a black-box model, based on a set of easily measurable exogenous inputs like pressures and temperatures at the stack and is able to predict the output voltage of the fuel cell stack. The model obtained is being exploited as a component of complex control systems able to manage the energy flows between fuel cell stack, battery pack, auxiliary systems and electric engine in a zero-emission vehicle prototype

In this paper the problem of regressor selection in virtual sensor design is addressed. In particular nonlinear models designed by experimental data are used to estimate relevant process variables of an industrial plant. The plant considered is a Sulphur Recovery Unit of a large refinery settled in Sicily. The proposed approach is used to face with the problem of input regressor selection of NMA models. The approach is based on a recursive evaluation of the cross correlation function between input variables and model residuals. The obtained results are compared with corresponding estimation obtained by using a reference model. Significant improvements in the model estimation capability show the suitability of the proposed method

This work deals with the discrimination of neutrons and gamma rays on the basis of their different pulse shapes in scintillator detectors; this technique is widely employed in nuclear fusion applications. After a thorough phase of data analysis, a multi layer perceptron (MLP) is trained with the aim of processing the shape of light pulses produced by these ionizing particles in an organic liquid scintillator and digitally acquired. Moreover, fast-superimposed events (called pile-ups) are detected and a further MLP is trained to analyze them and recover the original superimposed events. Satisfactory experimental results were obtained at the Frascati Tokamak Upgrade, ENEA-Frascati, Italy.

This paper gives a new contribution to characterize a class of complex systems build as arrays of coupled fuzzy logic based chaotic oscillators and investigates their dynamical features. Different spatio-temporal dynamics have been reproduced using interconnected fuzzy chaotic cells in order to study the effects, due to the variation of some parameters and network topologies, in the collective behavior and to highlight the synchronization capability of the complex fuzzy systems under consideration. The synchronization characteristics have been focused by defining a behavioral index.

In this paper, dynamical systems made up of locally coupled nonlinear units are used to control the locomotion of bio-inspired robots and, in particular, a simulation of an insect-like hexapod robot. These controllers are inspired by the biological paradigm of central pattern generators and are responsible for generating a locomotion gait. A general structure, which is able to change the locomotion gait according to environmental conditions, is introduced. This structure is based on an adaptive system, implemented by motor maps, and is able to learn the correct locomotion gait on the basis of a reward function. The proposed control system is validated by a large number of simulations carried out in a dynamic environment for simulating legged robots.

In this paper, we consider coupled quantum-dot cells, which are usually used for quantum-dot cellular automata, to build nanoscale dynamical systems. In particular, it is shown how the simple connection of few quantum-dot cells, quantum cellular nonlinear networks (Q-CNNs), can cause the onset of chaotic oscillations. Complex dynamics can be obtained only with small differences of polarizations and parameters. Local activity conditions are investigated for a two-cells case satisfying the criteria for the generation of complex spatio-temporal behaviors. The richness of dynamics of quantum CNNs is also emphasized through examples of synchronization in an array of so-built oscillators, in both cases of identical parameters and spatial dissymmetry.

Central Pattern Generators (CPGs) are a suitable paradigm to solve the problem of locomotion control in walking robots. CPGs are able to generate feed-forward signals to achieve a proper coordination among the robot legs. In literature they are often modelled as networks of coupled nonlinear systems. However the topic of feedback in these systems is rarely addressed. On the other hand feedback is essential for locomotion. In this paper the CPG for a hexapod robot is implemented through Cellular Neural Networks (CNNs). Feedback is included in the CPG controller by exploiting the dynamic properties of the CPG motor-neurons, such as synchronization issue and local bifurcations. These universal paradigms provide the essential issues to include sensory feedback in CPG architectures based on coupled nonlinear systems. Experiments on a dynamic model of a hexapod robot are presented to validate the approach introduced.

In this paper a CNNs based circuit for the generation of hyperchaotic signals is proposed. The circuit has been developed for applications in secure communication systems. An Saito oscillator has been designed by using a suitable configuration of a four-cells State-Controlled CNNs. A cryptography system based on the Saito oscillator has been implemented by using inverse system synchronization. The proposed circuit implementation and experimental results are given.

This paper proposes an experimental analysis on the convergence of evolutionary algorithms (EAs). The effect of introducing chaotic sequences instead of random ones during all the phases of the evolution process is investigated. The approach is based on the substitution of the random number generator (RNG) with chaotic sequences. Several numerical examples are reported in order to compare the performance of the EA using random and chaotic generators as regards to both the results and the convergence speed. The results obtained show that some chaotic sequences are always able to increase the value of some measured algorithm-performance indexes with respect to random sequences. Moreover, it is shown that EAs can be extremely sensitive to different RNGs. Some t-tests were performed to confirm the improvements introduced by the proposed strategy.

In this paper the problem of controlling the attitude of a rigid body, such as a Spacecraft, in three-dimensional space is approached by introducing two new control strategies developed in hypercomplex algebra. The proposed approaches are based on two parallel controllers, both derived in quaternion algebra. The first is a feedback controller of the proportional derivative (PD) type, while the second is a feedforward controller, which is implemented either by means of a hypercomplex multilayer perceptron (HMLP) neural network or by means of a hypercomplex radial basis function (HRBF) neural network. Several simulations show the performance of the two approaches. The results are also compared with a classical PD controller and with an adaptive controller, showing the improvements obtained by using neural networks, especially when an external disturbance acts on the rigid body. In particular the HMLP network gave better results when considering trajectories not presented during the learning phase.

In this paper the paradigm of Reaction-Diffusion Cellular Neural Networks, usually treated in bidimensional grids, is extended to three-dimensional spatial domains. In such a way it is possible to reproduce spatio-temporal dynamics met in spatio-temporal systems evolving in three-dimensional lattices. The peculiarities of the model, its realization by means of RD-CNNs, some simulations as well as the circuit implementation details are reported. Some interesting applications are also addressed.

In this paper it is shown that, with slight modifications, State Controlled CNNs (SC-CNNs) are able to approximate the behaviour of a class of complex dynamics with multivariable nonlinearities. In particular, in the so-called Extended SC-CNN defined in this work, the output nonlinearity shape has been modified, and a new template acting on the output function of the cell has been introduced. The needed circuitry to extend SC-CNNs, together with SPICE simulations of the new system, are here reported in order to confirm the suitability of the approach.

In this paper a new chaotic system based on two CNN-like cells is presented. Starting from the classical mathematical representation of Cellular Neural Networks, it is shown that the addition of a further parameter like the non-integer order m of each cell may cause the appearance of new complex behaviors. A suitable integer order approximation is therefore proposed and a chaotic circuit derived from it is realized.

The methodology proposed in the paper applies artificial intelligence (AI) techniques to the modeling and control of some climate variables within a greenhouse. The nonlinear physical phenomena governing the dynamics of temperature and humidity in such systems are, in fact, difficult to model and control using traditional techniques. The paper proposes a framework for the development of soft computing-based controllers in modern greenhouses.

A new type of micro walking robot named PLIF (piezo light intelligent flea) is introduced. These robots, that walk by using piezoceramic legs, are very small in size, but, at the same time, fast and agile. Three different types of PLIF have been designed and built and several dynamic measures have been performed. Moreover a self-learning technique has been implemented and tested in order to increase the autonomy of these systems.

In the paper a strategy in order to decrease the learning time without affecting the learning efficiency for a complex valued Multi Layer Perceptron (CMLP) is proposed. The methodology makes use of auxiliary devices that enable one to predict the connections trend of the principal neural network whose learning phase is not damaged since the auxiliary devices run concurrently with it. A numerical example is reported which shows the suitability of the proposed approach.

In this paper the behavior of an industrial flyback converter with current-mode control loop is discussed. A mixed electrical-functional model of an industrial flyback converter with current-mode control loop has been derived. Several experimental measurements on the VLSI implementation (VIPower) of the considered converter have been performed. The nonlinear behavior of the considered system has been investigated by using both numerical simulation and experimental analysis. Moreover a canonical "strange attractor" has been discovered.

In this paper the approximation capabilities of different structures of complex feedforward neural networks, reported in the literature, have been theoretically analyzed. In particular a new density theorem for Complex Multilayer Perceptrons with complex valued non-analytical sigmoidal activation functions has been proven. Such a result makes Multilayer Perceptrons with complex valued neurons universal interpolators of continuous complex valued functions. Moreover the approximation properties of superpositions of analytic activation functions have been investigated, proving that such combinations are not dense in the set of continuous complex valued functions. Several numerical examples have also been reported in order to show the advantages introduced by Complex Multilayer Perceptrons in terms of computational complexity with respect to the classical real MLP.

In the paper a new structure of a Multi-Layer Perceptron, able to deal with quaternion-valued signals, is proposed. A learning algorithm for the proposed Quaternion MLP (QMLP) is also derived. Such a neural network allows one to interpolate functions of a quaternion variable with a smaller number of connections with respect to the corresponding real valued MLP.>

In the paper it is proposed the use of a complex valued multi-layer perceptron neural network (MLP) with complex activation functions and complex connection strengths in order to perform the estimation of chaotic time series. In particular, the Ikeda map is taken into consideration. A comparison between the behavior of the real MLP and the complex one is also reported, showing that the complex valued MLP requires a smaller topology as well as a lower number of parameters in order to reach comparable performance.>

In this paper a new strategy for the design of fuzzy filters is proposed through the use of a genetic optimisation algorithm (GAs). The innovation of the proposed approach lies in the possibility to design fuzzy filters directly in the frequency domain. Comparisons between frequency responses of GAs fuzzy filters and traditional filters give satisfactory results, showing the validity of this technique.>

The problem of filtering tilt signals recorded in active volcanic areas is considered. The drawbacks of using traditional digital filtering techniques for this kind of data are outlined, and a model-based procedure is proposed. The model based procedure is based on three possible classes of models for the estimation of thermal effects on tilt data. The validity of this approach has been tested by using data recorded by an array operating on Mt. Etna (Italy). About eight months before an eruption occurred, the tilt component clearly started to rise, coming out form the confidence interval delimited by two horizontal rows.>

The implementation of an expert system making possible online fault diagnosis on a complex data acquisition system, a particle detector, designed for the MACRO (Monopoles, Astrophysics and Cosmic Ray Observatory) physics investigations is outlined. The system, called MADIES (Macro Diagnostic Expert System), is designed to take appropriate control actions in order to reduce the unavailability of the apparatus. MADIES was developed using the NEXPERT commercial shell.>