Artificial neural networks applications. Part 12 - The prediction of alkane heat capacities with the MolNet neural network

MolNet, a new multi-layer feedforward neural network ii; presented together with its application to the computation of alkane heat capacities. The Mol Net neural network changes its topology (the number pf neurons in the input and hidden layers, together with the number and type of connections) accor ding to the molecular structure of the chemical compound presented to the n etwork. The structure of each molecule is encoded in the corresponding mole cular graph that is used to set the MolNet topology. Three structural descr iptors derived from the molecular graph are used as input data for the firs t layer of neurons, namely the degree, the distance sum, and the reciprocal distance sum.