Deepnic

Deep learning based on convolutional neuron networks (CNN) is not suitable to classify non-image data.Instead of repositioning the input data to make an image, we have developed a new method for creating a specific image for each variable.This new method uses the results of Random Perfect Tree Forests, which is based on Nguyen Information Criterion (NIC).The values taken by these NICs under different conditions, make it possible to construct, for each variable, nine images. This method was successfully applied to the omic data (54675 variables) of the GSE22513 database concerning breast cancer resistance to placitaxel.The next step is to develop deep learning method for classifying these ROP-images.