Multilayer perceptron and vector quantization

The exponential encoding complexity has been the bottleneck drawback of vector quantization (VQ) in its applications. A kind of neural network multilayer perceptron (MLP) is introduced to attack the bottleneck. Based on the analysis of the VQ structure and the function of the MLP, an important conclusion on the relationship between the task and the scale required by it is drawn so that the two-layer MLP is adequate for VQ encoding or recognition. Simulation experiments are presented to test the theoretical analysis.<
>