Discussion on 'Correction of soil parameters in calculation of embankment settlement using a BP network back-analysis model' By Zhi-liang Wang, Yong-chi Li, R.F. Shen [Engineering Geology 91 (2007) 168–177]

The paper studied by Yong-chi Li and R.F. Shen (2007) includes some problems regarding the application of the proposed Neural Network (NN) approach. A very limited number of data has been used as testing set (only 3 among 52 data sets) which should be 20–30% of the database. The NN application in the study is not a function approximation problem where same input must always lead to the same output. NNs cannot be used for such an application. Moreover the generalization capability of the NN model has not been investigated. This discussion, aims to points out controversial points of the paper.