GMoretti@massey.ac.nz (Giovanni Moretti) (06/06/90)
>>> POINTERS TO IMPLEMENTING BP USING INTEGERS ONLY I'd like to use a neural net on a simple microprocessor (eg 8051) that has fast integer arithmetic but no (native) floating point. Therefore I'd like to use integer or maybe low precision floating point arithmetic. All the references I've seen are for algorithms using FP with lots of precision. Changing all REALS to INTEGERS with simple scaling doesn't work - I didn't expect it would but one lives in hope :-) Does anyone have any references to integer versions of the BP or similar algorithms or any thoughs on what precision is necessary (it'll probably depend on the number of neurons ...)? Thanks Giovanni -- ------------------------------------------------------------------------------- | GIOVANNI MORETTI, Consultant | EMail: G.Moretti@massey.ac.nz | |Computer Centre, Massey University | Ph 64 63 69099 x8398, FAX 64 63 505607 | | Palmerston North, New Zealand | QUITTERS NEVER WIN, WINNERS NEVER QUIT | -------------------------------------------------------------------------------