Hamming Distance Behaviour as an Analytical Tool of a Pattern Recogniser

Abstract

본 논문에서는 판정.피드백 루프를 가지는 디지탈학습망(Digital Learning Net)은 일정기간의 학습후, 미지의 외부입력에 대해 3싸이클의 상태공간 동작후에는 안정상태에 도달함을 보였다. 따라서 단층망군(Single-layer Nets)은 안정상태로부터 오차없이 외부 입력을 판정한다.

각 망의 응답은 해밍거리(Hamming Distance)에 의해 통계적으로 예측할 수 있음을 밝혔으며, 이 해석법은 미지의 시스템을 해??할 수 있는 강력한 방법을 제공한다.

In this paper, it is shown that after a period of training, The output of the digital learning net with decision feedback loops enters a stable state within three cycles of state space activity related to an unknown input pattern.

Thus the single-layer nets can make a strong decision from the preprocessed stable output.

Each response of the nets, both the digital learning net and the single-layer nets, is estimated in terms of Hamming distance behaviour, and this provides an important analytical tool for an unknown system.

In this paper, it is shown that after a period of training, The output of the digital learning net with decision feedback loops enters a stable state within three cycles of state space activity related to an unknown input pattern.

Thus the single-layer nets can make a strong decision from the preprocessed stable output.

Each response of the nets, both the digital learning net and the single-layer nets, is estimated in terms of Hamming distance behaviour, and this provides an important analytical tool for an unknown system.