(1) If your tables are small enough to be in RAM, store the characterizing parameters in eeprom and upon initialization recalculate your runtime tables.

(2) If your tables are too big for RAM but calculations are infrequent, store the tables in EEProm and access them serially when you infrequently calculate.

(3) If your tables are too big for RAM and too frequently calculated to rely upon serial EEProm access, you can either use external RAM with technique (1) or...

(4) Use in-system-programming FLASH to self-modify runtime tables.

If you use self-modifying code, you should use a spreadsheet to model your algorithm so you can uncover instabilities that may occur when its left upon its own.

duh