Fig. 2From: Convolutional neural network based on SMILES representation of compounds for detecting chemical motifOverview of our CNN. The SMILES string of a compound is represented as a feature matrix. CNN has multiple layers consisting of two convolutional and pooling layers with a subsequent global pooling layer. CNN is applied to the feature matrix and produces a low-dimensional feature representation (actually, 64-dimensional vector) termed the SCFP. Classification models are constructed by using SCFP as input for subsequent fully connected layersBack to article page