Asdasdasdasdasdasd
9420 palavras
38 páginas
SNNS Java VersionFunction Declaration File
Format: kernel name|presentation name|function type|no of parameters|1. parameter name|1. parameter tool tip text|1. parameter default value|2. parameter name|...
The function types are:
OUT_FUNC 1
ACT_FUNC 2
SITE_FUNC 3
LEARN_FUNC 4
UPDATE_FUNC 5
INIT_FUNC 6
ACT_DERIV_FUNC 7
JOG_WEIGHT_FUNC 8
ACT_2_DERIV_FUNC 9
PRUNING_FUNC 10
TEST_FUNC 11
REMAP_FUNC 12
PRUNING_LEARN 13
CC_LEARN 14
CC_MINI 15
CC_MODI 16
CC_ACT 17
You may use greek letters by unicode ( http:\\www.unicode.org ) for the parameter names and their tool tip texts.
You have to write them in the way: {0x03b1} for alpha, {0x03b2} for beta,...
Initializing functions:
Randomize_Weights|Random Weights|6|2|min|minimum value|-1.0|max|maximum value|1.0
ART1_Weights|for ART1 networks|6|2|{0x03b2}||1.0|{0x03b3}||1.0
ART2_Weights|for ART2 networks|6|2|d|Output value of F2 winner unit|0.9|{0x03b3}||1.2
ARTMAP_Weights|for ARTMAP networks|6|4|{0x03b2}a||1.0|{0x03b3}a||1.0|{0x03b2}b||1.0|{0x03b3}b||1.0
ClippHebb|ClippHebb|6|2|bias in||1.0|bias out||-1.0
CPN_Rand_Pat|for Counterpropagation, random pattern|6|0
CPN_Weights_v3.2|for Counterpropagation, ver. 3.2 (cube)|6|2|min|minimum value|-1.0|max|maximum value|1.0
CPN_Weights_v3.3|for Counterpropagation, ver. 3.3 (sphere)|6|2|min|minimum value|-1.0|max|maximum value|1.0
DLVQ_Weights|for Dynamic Learning Vector Quantization|6|2|min|minimum value|-1.0|max|maximum value|1.0
Hebb|Hebbian|6|2|bias in||1.0|bias out||-1.0
Hebb_Fixed_Act|Hebbian, fixed number of 1s|6|2|activity|number of active units of the patterns|10.0|distortion|degree of distortion in percent|10.0
JE_Weights|for Jordan or Elman networks|6|5|min|minimum