JavaDL
JavaDL is a java-based implementation of deep neural network (DNN) method. It has been used to predict cancer responses to drugs solely based on their chemical features. For the method, a novel cost function was devised by adding a regularization term to suppress overfitting. We also adopted an “early stopping” strategy to further reduce overfit and improve the accuracy and robustness of models. The software was also integrated with a genetic algorithm-based variable selection approach. For more information about its implementation and evaluation, please refer to our publication below.
Availability
The downloadable JavaDL is freely available here for non-commercial use. Currently, we are developing a new version, termed DL4DR, based on the latest convolutional neural network (CNN) techniques and big data of cancer genomics, as described here.