Performance evaluation of a high order data compressor using neural networks

A hierarchical neural network based data compressor/decompressor has been proposed, trained and evaluated by the authors. Further performance evaluation of this network is presented. The network was trained using the nested training algorithm (NTA). The hierarchical neural network architecture employing NTA training has been used successfully to compress image data. This network exhibited a high degree of parallelism, which aids in training. The method exhibited strong generalization capabilities over a wide class of images. In addition, the network showed robustness when hidden nodes are damaged