Re: Problems with Rprop-Algorithm in batchman-mode

• Messages sorted by: [ date ][ thread ][ subject ][ author ]
• Next message: : "re:SNNSv4.1 CCA & JE"
• Previous message: : "Re: cascade correlation in batch mode"
• Maybe in reply to: : "Problems with Rprop-Algorithm in batchman-mode"

hello Christoph, I looked into
 SNNSv4.1/kernel/sources/func_tbl.c
ans saw, that the JE functions

 {"JE_BP",               TEST_FUNC, 3, 1, (FunctionPtr) TEST_JE_Backprop},
 {"JE_BP_Momentum",      TEST_FUNC, 5, 1, (FunctionPtr) TEST_JE_BackpropMomentum},
 {"JE_Quickprop",        TEST_FUNC, 5, 1, (FunctionPtr) TEST_JE_BackpropMomentum},
 {"JE_Rprop",            TEST_FUNC, 4, 1, (FunctionPtr) TEST_JE_Rprop}, 

arn't commented. Therefore it have to be possible to stop training by
using batchman. If you don't know how to use batchman have a look at
 http://pikas.inf.tu-dresden.de/~trausch/index.html
If this isn't enough for you, mail us again and ask your concret questions.
ciao manfred

Manfred Weber; Carolo Wilhelmina - Technische Universitaet zu Braunschweig
Institut fuer Datenverarbeitungsanlagen, Haus der Elektrotechnik - 1205
Hans-Sommer-Strasse 66, D - 38106 Braunschweig, Germany
Phone +49\531\391\3746
email manfred@ida.ing.tu-bs.de

• Next message: : "re:SNNSv4.1 CCA & JE"
• Previous message: : "Re: cascade correlation in batch mode"
• Maybe in reply to: : "Problems with Rprop-Algorithm in batchman-mode"