krui.saveNet('foo.net','Testnet')
print krui.getLearnFunc()
#krui.setLearnFunc('RBF-DDA')
print krui.getInitialisationFunc()
print krui.getUpdateFunc()
patset = krui.loadNewPatterns('encoder.pat')
print "Patternset", patset, "loaded."
print "Old pattern number:", krui.getPatternNo()
# fiddle around with some patterns
krui.setPatternNo(8)
krui.deletePattern()
krui.setPatternNo(5)
krui.modifyPattern()
krui.setPatternNo(7)
krui.showPattern(krui.OUTPUT_OUT)
krui.newPattern()
krui.shufflePatterns(1)
krui.shuffleSubPatterns(1)
krui.saveNewPatterns('foo.pat',patset)
krui.setPatternNo(1)
(setinfo, patinfo) = krui.GetPatInfo()
print "Number of patterns: ", setinfo.number_of_pattern
print "Input Dimensions:", len(patinfo.input_dim_sizes)
print "Output Dimensions:", len(patinfo.output_dim_sizes)
print "Number of subpatterns for each pattern:", krui.DefTrainSubPat((),(),(),())
print "Total number of subpatterns:", krui.getTotalNoOfSubPatterns()
krui.setRemapFunc("None",())
print "Learning all patterns, result:"
print krui.learnAllPatterns((1.2,0))
print "Learning pattern 3, result:"
krui.saveNet('foo.net', 'Testnet')
print krui.getLearnFunc()
#krui.setLearnFunc('RBF-DDA')
print krui.getInitialisationFunc()
print krui.getUpdateFunc()
patset = krui.loadNewPatterns('encoder.pat')
print "Patternset", patset, "loaded."
print "Old pattern number:", krui.getPatternNo()
# fiddle around with some patterns
krui.setPatternNo(8)
krui.deletePattern()
krui.setPatternNo(5)
krui.modifyPattern()
krui.setPatternNo(7)
krui.showPattern(krui.OUTPUT_OUT)
krui.newPattern()
krui.shufflePatterns(1)
krui.shuffleSubPatterns(1)
krui.saveNewPatterns('foo.pat', patset)
krui.setPatternNo(1)
(setinfo, patinfo) = krui.GetPatInfo()
print "Number of patterns: ", setinfo.number_of_pattern
print "Input Dimensions:", len(patinfo.input_dim_sizes)
print "Output Dimensions:", len(patinfo.output_dim_sizes)
print "Number of subpatterns for each pattern:", krui.DefTrainSubPat(
(), (), (), ())
print "Total number of subpatterns:", krui.getTotalNoOfSubPatterns()
krui.setRemapFunc("None", ())
print "Learning all patterns, result:"
print krui.learnAllPatterns((1.2, 0))
#!/usr/bin/python
# shows the coordinates of the winner neurons for the som_cube example
from snns import krui, util
krui.loadNet('som_cube.net')
krui.loadNewPatterns('som_cube.pat')
patnum = krui.getNoOfPatterns()
units = krui.getNoOfUnits()
for pat in range(1,patnum+1) :
krui.setPatternNo(pat)
krui.showPattern(krui.OUTPUT_NOTHING)
krui.updateNet(())
results = []
for unit in range(1,units+1) :
if krui.getUnitTType(unit) == krui.HIDDEN :
results.append((krui.getUnitActivation(unit),unit))
bestact, bestunit = min(results)
rawpos = krui.getUnitPosition(bestunit)[:2]
print "Pattern", pat, "Act", bestact, "Unit", bestunit,
print "Grid Position", (rawpos[0]-4, rawpos[1])
i = i + 1
# druhá fáze (jemnìj¹í)
i=0
while i < pruch_2:
res = krui.learnAllPatterns(0.03, 0.1)
if not i % 100 : print "Fáze 2, chyba v cyklu %d:" % i, res[0]
i = i + 1
print "Rekonstruuji pùvodní obrázek"
im.blank()
for p in xrange(krui.getNoOfPatterns()):
krui.setPatternNo(p + 1)
krui.showPattern(1);
krui.updateNet()
block = []
for u in xrange(64 + 16 + 1, 64 * 2 + 16 + 1):
block.append(real2pix(krui.getUnitActivation(u)))
im.setBlock(p, block)
print "Zapisuji výsledný obrázek na disk"
im.store("compressed.rgb")
print "Vytváøím soubor vzorkù pro SNNS"
krui.saveNewPatterns('image.pat', patset)
print "Vytváøím soubor sítì pro SNNS"
krui.saveNet('image.net','image')