Learning and convergence analysis of neural-type structurednetworks

A class of feedforward neural networks, structured networks, has recently been introduced as a method for solving matrix algebra problems in an inherently parallel formulation. A convergence analysis for the training of structured networks is presented. Since the learning techniques used in structured networks are also employed in the training of neural networks, the issue of convergence is discussed not only from a numerical algebra perspective but also as a means of deriving insight into connectionist learning. Bounds on the learning rate are developed under which exponential convergence of the weights to their correct values is proved for a class of matrix algebra problems that includes linear equation solving, matrix inversion, and Lyapunov equation solving. For a special class of problems, the orthogonalized back-propagation algorithm, an optimal recursive update law for minimizing a least-squares cost functional, is introduced. It guarantees exact convergence in one epoch. Several learning issues are investigated.     

Opposite-channel-based charge injection in SOI MOSFET's under hotcarrier stress

By operating one channel of a typical SOI MOSFET in avalanche while keeping the opposite channel accumulated, charge injection into the opposite gate takes place. Three independent experiments are described that demonstrate the occurrence of this opposite-channel based charge injection. The experimental results are in agreement with PISCES numerical simulations     

A Glance into Social and Evolutionary Aspects of an Artifact Ecology for Collaborative Learning through the Lens of Distributed Cognition

There is steadily growing interest in the design and evaluation of digitally enhanced spaces in the real-world where users use, combine, and appropriate different physical and digital tools based on the task at hand. Under the notion of “ubiquitous computing,” researchers have explored tools and theories to analyze, understand, and model users in complex socio-technical systems. Yet, even though theoretical approaches play a significant role in the work of practitioners, there is a general difficulty in applying them, pushing researchers to explore methodological frameworks with clearer guidelines. Using the distributed cognition for teamwork (DiCoT) framework, we study collaboration and communication patterns, physical movement, and social structures of two groups of learners working on a design problem as they evolve over a 3-month period. Through an in-depth investigation, we present detailed accounts of the social and evolutionary models of DiCoT for each group. Our analysis enriches the DiCoT framework by identifying five new principles: social emersion and social circles of privacy for the social model, and continuity, mutual adaptation, and semantics of body for the evolutionary model. This article contributes to HCI research by refining and extending the existing DiCoT framework and elaborating on two dimensions that have previously been under-developed in the literature.     

A review of current research in manufacturing shop design integration

The importance of the manufacturing shop design in the successful operation of a production system is well documented and, as a result, significant work has been devoted to this area. Due to the inherent complexity of the global shop design problem, much research has focused on individual sub-problems, including the layout of manufacturing resources, the design of the shop aisle network, and the selection of the material handling equipment, rather than addressing the integrated problem directly. This paper reviews literature contributions to shop design integration, as well as partial integration schemes such as the design for operation and control of material handling systems. Furthermore, it examines the ability of models and solution approaches for individual sub-problems of shop design to support integration schemes.     

Directed Evolution of Phi Class Glutathione Transferases Involved in Multiple-Herbicide Resistance of Grass Weeds and Crops

The extensive application of herbicides in crop cultivation has indisputably led to the emergence of weed populations characterized by multiple herbicide resistance (MHR). This phenomenon is associated with the enhanced metabolism and detoxifying ability of endogenous enzymes, such as phi class glutathione transferases (GSTFs). In the present work, a library of mutant GSTFs was created by in vitro directed evolution via DNA shuffling. Selected gstf genes from the weeds Alopecurus myosuroides and Lolium rigidum, and the cereal crops Triticum durum and Hordeum vulgare were recombined to forge a library of novel chimeric GSTFs. The library was activity screened and the best-performing enzyme variants were purified and characterized. The work allowed the identification of enzyme variants that exhibit an eight-fold improvement in their catalytic efficiency, higher thermal stability (8.3 °C) and three-times higher inhibition sensitivity towards the herbicide butachlor. The crystal structures of the best-performing enzyme variants were determined by X-ray crystallography. Structural analysis allowed the identification of specific structural elements that are responsible for k_cat regulation, thermal stability and inhibition potency. These improved novel enzymes hold the potential for utilization in biocatalysis and green biotechnology applications. The results of the present work contribute significantly to our knowledge of the structure and function of phi class plant GSTs and shed light on their involvement in the mechanisms of MHR.     

Frozen mirabelle plum drying: Kinetics, modelling and impact on biochemical properties

The aim of this research was to study and model the kinetics of the hot air drying of frozen mirabelle plums. Effects of temperature (50-85 °C), air velocity (0.6-1.2 m/s) and a pre-treatment (ascorbic acid + sucrose-saturated solution) were investigated. The sorption isotherm and heat of sorption of mirabelle plum were mathematically described respectively by the Guggenheim, Anderson and de Boer (GAB) model and a new one. Depending on temperature and air velocities, the effective diffusivity and the energy of activation were respectively in the range of 8-16.8 × 10⁻¹⁰ m² s⁻¹ and 43.7-55.14 kJ mol⁻¹. To establish the model best adapted to describe this process the fit quality of five models (Newton, page, modified page, logarithmic and diffusion) were evaluated and compared. The diffusion model led to low values of reduced χ-square, mean square root and mean relative deviation modulus with the highest linear regression coefficient. This model was validated in a large range of operating conditions and represents an excellent tool to predict the duration of this process. The pre-treatment preserves biochemical properties of the dried mirabelle plums (reducing sugar content, total phenolic compounds and anti-oxidant activity).     

Robust switching adaptive control of multi-input nonlinear systems

During the last decade a considerable progress has been made in the design of stabilizing controllers for nonlinear systems with known and unknown constant parameters. New design tools such as adaptive feedback linearization, adaptive back-stepping, control Lyapunov functions (CLFs) and robust control Lyapunov functions (RCLFs), nonlinear damping and switching adaptive control have been introduced. Most of the results developed are applicable to single-input feedback-linearizable systems and parametric-strict-feedback systems. These results, however, cannot be applied to multi-input feedback-linearizable systems, parametric-pure-feedback systems and systems that admit a linear-in-the-parameters CLF. In this paper, we develop a general procedure for designing robust adaptive controllers for a large class of multi-input nonlinear systems. This class of nonlinear systems includes as a special case multi-input feedback-linearizable systems, parametric-pure-feedback systems and systems that admit a linear-in-the-parameters CLF. The proposed approach uses tools from the theory of RCLF and the switching adaptive controllers proposed by the authors for overcoming the problem of computing the feedback control law when the estimation model becomes uncontrollable. The proposed control approach has also been shown to be robust with respect to exogenous bounded input disturbances     

Parameter convergence of a new class of adaptive controllers

A new class of adaptive control schemes for minimum-phase linear time invariant (LTI) systems has recently been developed using nonlinear design techniques which guarantee improved transient performance in addition to closed-loop stability and asymptotic tracking. In this paper we establish the parameter convergence properties of this new class of schemes in the presence of persistently exciting signals and compare them with the properties of the traditional adaptive controllers. We show that the new class of adaptive controllers has stronger parameter convergence properties in the presence of overparameterization     

High-order neural network structures for identification ofdynamical systems

Several continuous-time and discrete-time recurrent neural network models have been developed and applied to various engineering problems. One of the difficulties encountered in the application of recurrent networks is the derivation of efficient learning algorithms that also guarantee the stability of the overall system. This paper studies the approximation and learning properties of one class of recurrent networks, known as high-order neural networks; and applies these architectures to the identification of dynamical systems. In recurrent high-order neural networks, the dynamic components are distributed throughout the network in the form of dynamic neurons. It is shown that if enough high-order connections are allowed then this network is capable of approximating arbitrary dynamical systems. Identification schemes based on high-order network architectures are designed and analyzed.