A robust neural model - Modified Learning Vector Quantization (MLVQ) - is proposed for the estimation of centroid in pattern recognition. This MLVQ model can significantly demonstrate the behavior of better estimating the class centroid by utilizing the distance-dependent step size. The computer simulation result can indicate the high potential of less dependence on the initial point as well as the precise settlement of the weight vectors to the centroids. The main feature of this model is robust to the noise perturbation to the pattern distributions in practical applications. Also, a mixed-mode of analog-digital processing systems are designed by the CMOS current-mode VLSI technology and offer the best attributes of both analog and digital computation. The final experimental results of this hybrid processing systems can show the on-chip learning capability and operate at microsecond time scale to achieve the goal of real-time neural applications.
|Number of pages||4|
|Journal||Proceedings - IEEE International Symposium on Circuits and Systems|
|State||Published - 1 Jan 1996|
|Event||Proceedings of the 1996 IEEE International Symposium on Circuits and Systems, ISCAS. Part 1 (of 4) - Atlanta, GA, USA|
Duration: 12 May 1996 → 15 May 1996