|Structure Input||3D-Structure Generation||Calculation of Properties||Structure Coding||Network Training||IR Spectrum Simulation|
A trained neural network is able to predict an infrared spectrum for a molecule it has not seen before. The structure code of the query structure is compared with the structure block of the neural network (red). The similarity criterion is the rms error between the query structure code and the neuron weights. The most similar neuron, the winning neuron, acts as a poniter to the corresponding information in the output block (green), The output block provides the simulated spectrum.