A Learning System Which Accommodates Feature Interactions

The a u t h o r ' s s ta te -space l e a r n i n g system has e f f e c t i v e l y op t im ized the c o e f f i c i e n t s o f l i n e a r e v a l u a t i o n f u n c t i o n s . The incrementa l approach uses s t a t i s t i c a l performance measures from completed s o l u t i o n s to boo t s t r ap the h e u r i s t i c , which es t imates p r o b a b i l i t y o f task u s e f u l n e s s . These s t a t i s t i c s are c l u s t e r e d in f ea tu re space, forming a media t ing knowledge s t r u c t u r e ( reg ion se t ) between the d i r e c t performance measures and the genera l i zed e v a l u a t i o n f u n c t i o n . The reg ions are da ta -de te rm ined , i n s e n s i t i v e to n o i s e , and a l l ow management o f i n t e r a c t i n g f ea tu res through n a t u r a l p iecewise l i n e a r i t y . Ear ly exper iment w i t h n o n l i n e a r i t y i n d i c a t e s s t a b i l i t y , f l e x i b i l i t y and improved task per formance.