Search

In this paper, a new pruning strategy based on the neuroplasticity of biological neural networks is presented. The novel pruning algorithm proposed is inspired by the knowledge remapping ability after injuries in the cerebral cortex. Thus, it is proposed to simulate induced injuries into the network by pruning full convolutional layers or entire blocks, assuming that the knowledge from the removed segments of the network may be remapped and compressed during the recovery (retraining) process. To reconnect the remaining segments of the network, a translator block is introduced. The translator is composed of a pooling layer and a convolutional layer. The pooling layer is optional and placed to ensure that the spatial dimension of the feature maps matches across the pruned segments. After that, a convolutional layer (simulating the intact cortex) is placed to ensure that the depth of the feature maps matches and is used to remap the removed knowledge. As a result, lightweight, efficient and accurate sub-networks are created from the base models. Comparison analysis shows that in our approach is not necessary to define a threshold or metric as the criterion to prune the network in contrast to other pruning methods. Instead, only the origin and destination of the prune and reconnection points must be determined for the translator connection. [ABSTRACT FROM AUTHOR]

This work proposes a discrete-time non-linear neural observer based on a recurrent high order neural network in parallel model trained with an algorithm based on the extended Kalman filter for discrete-time multiple input multiple output non-linear systems with unknown dynamics and unknown time-delay. To prove the semi-globally uniformly ultimately boundedness of the proposed neural observer the stability analysis based on the Lyapunov approach is included. Applicability of the proposed observer is shown via simulation and experimental results. [ABSTRACT FROM AUTHOR]

This work proposes a discrete-time nonlinear neural identifier based on a recurrent high-order neural network trained with an extended Kalman filter-based algorithm for discrete-time deterministic multiple-input multiple-output systems with unknown dynamics and time-delay. To prove the semi-globally uniformly ultimately boundedness of the proposed neural identifier, the stability analysis based on the Lyapunov approach is included. Applicability of the proposed identifier is shown via simulation and experimental results, all of them performed under the presence of unknown external and internal disturbances as well as unknown time-delays. [ABSTRACT FROM AUTHOR]

This paper proposes a neural control integrating stereo vision feedback for driving a mobile robot. The proposed approach consists in synthesizing a suitable inverse optimal control to avoid solving the Hamilton Jacobi Bellman equation associated to nonlinear system optimal control. The mobile robot dynamics is approximated by an identifier using a discrete-time recurrent high order neural network, trained with an extended Kalman filter algorithm. The desired trajectory of the robot is computed during navigation using a stereo camera sensor. Simulation and experimental result are presented to illustrate the effectiveness of the proposed control scheme. [ABSTRACT FROM AUTHOR]

We present an inverse optimal neural controller for a nonholonomic mobile robot with parameter uncertainties and unknown external disturbances. The neural controller is based on a discrete-time recurrent high order neural network (RHONN) trained with an extended Kalman filter. The reference velocities for the neural controller are obtained with a visual sensor. The effectiveness of the proposed approach is tested by simulations and real-time experiments. [ABSTRACT FROM AUTHOR]

This paper presented an inverse optimal neural controller with speed gradient (SG) for discrete-time unknown nonlinear systems in the presence of external disturbances and parameter uncertainties, for a power electric system with different types of faults in the transmission lines including load variations. It is based on a discrete-time recurrent high order neural network (RHONN) trained with an extended Kalman filter (EKF) based algorithm. It is well known that electric power grids are considered as complex systems due to their interconections and number of state variables; then, in this paper, a reduced neural model for synchronous machine is proposed for the stabilization of nine bus system in the presence of a fault in three different cases in the lines of transmission. [ABSTRACT FROM AUTHOR]

The ability to obtain accurate volatility forecasts is an important issue for the financial analyst. In this paper, we use the DAN2 model, a multilayer perceptron and an ARCH model to predict the monthly conditional variance of stock prices. The results show that DAN2 model is more accurate for predicting in-sample and out-of-sample variance that the other considered models for the used dataset. Thus, the value of this neural network as a predictive tool is demonstrated. [ABSTRACT FROM AUTHOR]

Time series prediction is an important research problem due to its implications in engineering, economics, finance and social sciences. An important topic about this problematic is the development of new models and its comparison with previous approaches in terms of forecast accuracy. Recently, support vector machines (SVM) have been used for time series prediction, but the reported experiences are limited and there are some problems related to its specification. The aim of this paper is to propose a novel technique for estimating some constants of the SVM usually fixed empirically by the modeler. The proposed technique is used to estimate several SVM with the aim of forecast five benchmark time series; the obtained results are compared with the statistics reported in other papers. The proposed method allow us to obtain competitive SVM for the time series forecasted in comparison with the results obtained using other most traditional models. [ABSTRACT FROM AUTHOR]

This work presents a real-time discrete nonlinear neural identifier based on a Recurrent High Order Neural Network (RHONN) trained online with an Extended Kalman Filter (EKF) based algorithm applied to a Linear Induction Motor (LIM). For the obtained neural model, a discrete-time sliding mode control law is designed for trajectory tracking of velocity and flux magnitude. The stability analysis is also included, based on the Lyapunov approach. This work is implemented in real-time by using MATLAB ® , 1 1 MATLAB is a registered trademark of The MathWorks, Inc. a dSPACE ® 2 2 dSPACE is a registered trademark of DSPACE GmbH. DS1104 controller board and its software RTI libraries and ControlDesk ® , 3 3 ControlDesk is a registered trademark of DSPACE GmbH. respectively, to control a Linear Induction Motor Lab - Volt ® 4 4 LAB-Volt is a registered trademark of Lab-Volt Systems, Inc. 8228. [ABSTRACT FROM AUTHOR]