Extended Data Fig. 8: Learned weights with training patterns 3 and 4 or 6 and 7.
a, Overlaid training patterns (10 per class) for learning handwritten digits 3, 4, 6 and 7. b,c, Fluorescence kinetics data, endpoint signal values at 4 hours, and error statistics for learning 3 first into memory 2 and then 4 into memory 1 (b) or learning 6 first into memory 2 and then 7 into memory 1.
