The non-invariance of pattern recognition measures using neural networks as applied to the protein secondary structure prediction problem

In this supervised pattern classification quandary, solutions to the protein secondary structure prediction problem using various neural network architectures are considered. Upon constructing mappings between training vectors and their desired targets, the class membership of test data and associated measures of significance have been found to vary depending on the set of applied target vectors. In this work, the linear neural network (LNN) and the backpropagation neural network (BPNN) are trained and tested using an existing database of 513 proteins. It is demonstrated that, on average, although individual classification results vary with the set of target vectors, overall recognition results remain consistent with those in the literature.