One of the most recent modifications on Widrow and Hoff's LMS algorithm has been the inclusion of a momentum term into the weight update equation. The resulting algorithm is referred to as “The Momentum LMS (MLMS...
详细信息
One of the most recent modifications on Widrow and Hoff's LMS algorithm has been the inclusion of a momentum term into the weight update equation. The resulting algorithm is referred to as “The Momentum LMS (MLMS) algorithm”. This paper revises the basic properties of the MLMS algorithm for stationary inputs. As a result, new bounds, on the parameters of the algorithm, for convergence are found, and it is shown that, under slow convergence conditions, this new algorithm is equivalent to the usual LMS algorithm, but it outperforms the LMS algorithm for fast convergence cases and for inputs containing inpulsive noise components. Zusammenfassung Eine der neuesten Modifikationen des “Kleinste-Quadrate-(LMS) algorithmus” nach Widrow und Hoff war die Einführung eines Momenten-Terms in die Beziehung, mit der die Gewichte aktualisiert werden. Der so entstandene algorithmus wird “Momenten-LMS-(MLMS) algorithmus” gennant. Im folgenden Beitrag werden die Grundeigenschaften des MLMS-Verfahrens für stationäre Signale neu dargestellt. Das führt zu neuen Konvergenz-Grenzwerten für die Parameter des algorithmus'. Es wird gezeigt, daβ das Verfahren unter den Bedingungen langsamer Konvergenz der gewöhnlichen LMS-Methode gleichwertig ist, im Falle schneller Konvergenz sowie bei Signalen mit Impulsstörungen jedoch besser arbeitet.
Anti-Freezing Asphalt Pavement (AFAP) has good snow-melting performance and is used widely in many countries around the world. The objective of this study was to analyze AFAP's long-term performance and predict it...
详细信息
Anti-Freezing Asphalt Pavement (AFAP) has good snow-melting performance and is used widely in many countries around the world. The objective of this study was to analyze AFAP's long-term performance and predict its snow-melting ability. Two types of anti-freezing stone matrix asphalt (SMA) mixtures (SMA-13 with Iceguard and SMA-13 with Mafilon) were prepared with the Marshall method. Water stability, high-temperature stability, low-temperature crack resistance, and freeze-thaw split tests were conducted to evaluate mixtures' performance. Meanwhile, the effect of anti-freezing filler, asphalt content, and soaking temperature on the salt dissolution of anti-freezing asphalt mixtures was analyzed, and the snow-melting ability of AFAP was predicted based on the backpropagation (BP) neural network. The results illustrated that water stability of anti-icing asphalt mixture reduced, and the dynamic stability after short-term aging was improved. The tensile strain and tensile strength ratio of the anti-icing asphalt mixture reduced after long-term aging and soaking in water. In addition, the salt dissolution rate increased with the increase of anti-freezing filler content and the decrease of asphalt content. The research conducted suggests that the BP Neural Network can be utilized to predict the snow-melting ability of the anti-freezing asphalt mixture, and the regression coefficient of the predicted and measured salt dissolution was higher.
Artificial neural networks have been employed in diverse applications ranging from control, to pattern recognition and classification. While password detection can be implemented with a digital electronic circuit with...
详细信息
ISBN:
(纸本)9781424433964
Artificial neural networks have been employed in diverse applications ranging from control, to pattern recognition and classification. While password detection can be implemented with a digital electronic circuit with non-volatile memory, this implementation is prone to hacking. In this paper, we present a Slayer feedforward neural network which we have designed, trained and tested for secure password detection. This network correctly detects with 100% accuracy when a password presented at its inputs matches its associated user ID which the network memorized during the training phase. The BrainMaker and NetMaker tools were used for training, simulating and testing our neural network. The paper also reports our experiments with increasing the number of hidden layers, number of neurons in the hidden layer, and noise additions on the network's detection accuracy.
As computer networks are grows exponentially security in computer system has become a foremost issue. Monitoring atypical activity can be one way to detect any violation that impedes computer systems security. Existin...
详细信息
ISBN:
(纸本)9781424420148
As computer networks are grows exponentially security in computer system has become a foremost issue. Monitoring atypical activity can be one way to detect any violation that impedes computer systems security. Existing methods like statistical models [12] for intrusion detection not perform well whereas Neural network has been proved as an efficient method for intrusion detection [10]. In this paper Feed-forward and Recurrent Neural network is trained by back propagation training algorithm and using normal data. Performances of these Neural Networks are compared against both normal data and intrusive data.
暂无评论