ALL-OPTICAL IMPLEMENTATION OF A SELF-ORGANIZING MAP - LEARNING AND TAXONOMY CAPABILITY ASSESSMENT

We present some experimental results obtained with an all-optical impl ementation of a self-organizing map. The system is based on two ferroe lectric liquid-crystal, bistable optically addressed spatial light mod ulators (BOASLM's) used as parallel read/write optical memories and as thresholding devices. Modifications made to adapt the basic algorithm to the BOASLM's are studied in computer simulations that predict trai ning-class loss due to the simplified decision stage as the principal drawback. Experimental results show good system behavior in recognitio n mode, demonstrating generalization properties and confirming the cho ice of a BOASLM memory. All-optical learning is shown in simple cases, and the elementary BOASLM decision function is confirmed as the major limiting factor. The ultimate possible system performance is discusse d, and ways of overcoming present limits are suggested. (C) 1995 Optic al Society of America