ABSTRACT
Characterization of porous materials is an attractive topic in the applied research studies. Efficient techniques are required to predict proper values of characterization parameters for the porous material. A novel method is introduced in the present article based on a special class of neural network known as Regularization network. A reliable procedure is presented for efficient training of the optimal network using two experimental data sets on characterization of activated carbon and carbon molecular sieve (CMS). These case studies were employed to compare the performances of two properly trained Regularization networks with conventional methods. It is also demonstrated that such Regularization networks provide more appropriate generalization performance over the conventional techniques.
