Skip to main content
Fig. 10 | Journal of Cheminformatics

Fig. 10

From: Chemoinformatics and structural bioinformatics in OCaml

Fig. 10

Git diff excerpt of an actual code refactoring in the SVM part of a category-QSAR software. Sometimes, the R svmpath package encounters numerical problems, like an exactly or computationally singular matrix. To deal with such rare cases, it was decided to drop a model from the bag of models. Since a bagging classifier with 21 models was being trained, dropping one or two models was deemed better than letting the whole software crash. Hence, an option type was introduced in the function optimal_lambda from file ‘svm_common.ml’, along with proper warning messages. Then, the compiler forced updating the rest of the code

Back to article page