In this paper, we propose a new learning-guided approach for diagnosis of intra-cell defects that may occur in customer returns. In the first part of the paper, only static defects modeled by stuck-at faults have been assumed. Several supervised learning algorithms were considered, with various levels of efficiency. In the second part of the paper, we have extended the previous work by dealing with more sophisticated (i.e. dynamic) defects. This time, we concentrated on a Bayesian classification method used for predicting the nature (likelihood to be a good candidate) of each new data instance (defect) that has to be evaluated during the diagnosis process. Results obtained on benchmark circuits, and comparison with a commercial cellaware diagnosis tool, demonstrate the efficiency of the proposed approach in terms of accuracy and resolution.