Biological pattern recognition by neural networks

Neural network analysis was proposed and evaluated as a method for image analysis of plankton data derived from automatic counting techniques. It was shown that a neural network with 2 layers of weights was capable of learning a large data set by the backward-error propagation method. Significant results were achieved in separating novel images of 2 co-occurring species of Ceratium spp. from the western North Atlantic Ocean.