Gustavo Guimarães Parma

Training neural networks with a multi-objective sliding mode control algorithm

Abstract This paper presents a new sliding mode control algorithm that is able to guide the trajectory of a multi-layer perceptron within the plane formed by the two objective functions: training set error and norm of the weight vectors. The results show that the neural networks obtained are able to generate an approximation to the Pareto set, from which an improved generalization performance model is selected.

On-line neural training algorithm with sliding mode control and adaptive learning rate

This paper presents a new algorithm for on-line artificial neural networks (ANN) training. The network topology is a standard multilayer perceptron (MLP) and the training algorithm is based on the theory of variable structure systems (VSS) and sliding mode control (SMC). The main feature of this novel procedure is the adaptability of the gain (learning rate), which is obtained from sliding mode surface so that system stability is guaranteed.

Neural networks learning with sliding mode control: the sliding mode backpropagation algorithm.

Based on the classical backpropagation weight update equations, sliding mode control theory is introduced as a technique to adapt weights of a multi-layer perceptron. As will be demonstrated, the introduction of sliding mode has resulted in a much faster version of the standard backpropagation. The results show also that the proposed algorithm presents some important features of sliding mode control, which are robustness and high speed of learning. In addition to that, this paper shows also how control theory can be applied to train neural networks.

Improving backpropagation with sliding mode control

Sliding mode control is applied as a procedure to adapt weights of a multilayer perceptron. Standard backpropagation weight update equations are used for providing error estimates for the output and hidden layers, similarly to the classical algorithm. The sliding mode procedures are then introduced to adapt weights taking into consideration the standard backpropagation errors. As demonstrated throughout this paper, the introduction of sliding mode has resulted in a much faster version of the standard backpropagation. The speed-up achieved is around two times the standard version.

Real-Time Digital Hardware Simulation of Power Electronics and Drives

This paper presents a digital hardware realization of a real-time simulator for a complete induction machine drive using a field-programmable gate array (FPGA) as the computational engine. The simulator was developed using Very High Speed Integrated Circuit Hardware Description Language (VHDL), making it flexible and portable. A novel device-characteristic based model suitable for FPGA implementation has been proposed for the 2-level 6-pulse IGBT-based voltage-source converter (VSC). The VSC model is computed at a fixed time-step of 12.5 ns allowing a highly detailed and precise accounting of gating signals. The simulator also models a squirrel cage induction machine, a direct field-oriented control system, a space-vector pulse-width modulation scheme (SVPWM) and a measurement system. A multirate simulation of the system shows the slow (machine) as well as the fast (VSC and control) dynamic components. Real time simulation results under steady-state and transient conditions demonstrate modeling accuracy and efficiency

Sliding mode backpropagation: control theory applied to neural network learning

This paper shows two different methodologies, both based on sliding mode control to train multilayer perceptron. These two methods are compared with standard back propagation, momentum and RPROP algorithms. The results show that the use of this control theory can reduce the time to train multilayer perceptron and also provide an interesting tool to analyze the limits for the parameters involved in the algorithm.

On-line training algorithms for an induction motor stator flux neural observer

This work presents a neural network based stator flux observer. Although the network topology is a standard multilayer perceptron network, the training algorithms are new. This paper presents two on-line training algorithms, which are based on Variable Structure Systems (VSS) theory and Sliding Mode Control (SMC). The resulting observer shows good convergence velocity and robustness with respect to the induction motor parameters for both training algorithms tested.

Control of generalization with a bi-objective sliding mode control algorithm

This paper presents a new sliding mode control algorithm that is able to guide the trajectory of a multilayer perceptron within the plane formed by the two objectives: training set error and norm of the weight vectors. The results show that the neural networks obtained are able to generate the Pareto set, from which a model with the smallest validation error is selected.

A New Neurofuzzy Controller Based on NFN Networks

This work presents a new online learning controller, the ONFC (Online Neurofuzzy Controller), which has as base the Neo Fuzzy Neuron – NFN. Its principal difference from the most of the neurofuzzy structure used in control systems is the fact that the process error is not only used to correct the network parameters, but also as network input. Moreover, the ONFC has a very simple structure with only one input and one output, associated by two fuzzy rules. Other important characteristic presented by this controller is the reduced effort for the fixed parameters adjustment. The proposed controller development is presented for single and multi-loop control. This controller is applied for the control of two different plants. In a single loop control, simulations results are obtained for a generic plant with reverse characteristic. In a multi-loop control, the controller performance is evaluated through a practical implementation of an induction motor vector control with stator field orientation.