Analysis of learning rate and momentum term in backpropagation neural network algorithm trained to predict pavement performance

You are here: Homepage > Search

Analysis of learning rate and momentum term in backpropagation neural network algorithm trained to predict pavement performance

Attoh-Okine, N.O.

Pavement performance modeling is a critical component of any pavement management system (PMS) decision-making process. A characteristic feature of pavement performance models is that they are formulated and estimated statistically from field data. The statistical modeling can consider no more than a few of the parameters, in a simplified manner, and in some cases various transformations of the original data. Lately, artificial neural networks (ANNs) have been applied to pavement performance modeling. ANNs offer a number of advantages over the traditional statistical methods, caused by their generalization, massive parallelism and ability to offer real-time solutions. Unfortunately, in previous pavement performance models, only simulated data were used in the ANN environment. In this paper, real pavement conditions and traffic data and a specific architecture are used to investigate the effect of the learning rate and the momentum term on a backpropagation algorithm neural network trained to predict flexible pavement performance. On the basis of the analysis, it is concluded that an extremely low learning rate (around 0.001-0.005) and momentum term (between 0.5 and 0.9) do not give satisfactory results for the specific data set and architecture used. It is also established that the learning rate and momentum term, together with validation data can be used to identify when over-learning is taking place in a training set.