CHEMNET - A NOVEL NEURAL-NETWORK-BASED METHOD FOR GRAPH PROPERTY MAPPING

ChemNet is a new method for mapping molecular properties. The input of ChemNet consists of two-dimensional matrices of variable sizes, inste ad of the sets of molecular descriptors used by the conventional mappi ng methods. The two-dimensional matrices considered in this study are graph distance matrices. The diagonal elements of the matrices are ato mic properties. ChemNet uses these matrices to form the topology of th e artificial neural network. Each molecule of a training set correspon ds to a single network configuration. The weighted connections of the networks are adjusted, using the ''backprop'' procedure. The original background of the method and details of the current realization are pr esented. Examples of how ChemNet learns topological and physicochemica l molecular properties demonstrate the practical use of the method.