Science Daily January 30, 2019
An international team of researchers (Finland, Denmark, Germany) has developed and trained the architecture ARTIST, made of three different neural network architectures: multilayer perceptron (MLP), convolutional neural network (CNN), and deep tensor neural network (DTNN), to predict molecular excitation using the coordinates and charges of the constituent atoms of each molecule as input. In demonstration they scanned the structures of 10k previously unseen organic molecules and obtained their spectra predictions to identify molecules for potential applications. The research has the potential to speed up the development of flexible electronics, produce better batteries and catalysts and create new compounds. ARTIST is currently available for use and further training upon request…read more. Open Access TECHNICAL ARTICLE