[conv1a]
epsW=0.01
epsB=0.02
momW=0.9
momB=0.9
wc=0.0005
wball=0.00
schedW=jdexp[250,0,4]
schedB=dexp[10,0,2]

[conv1b]
epsW=0.01
epsB=0.02
momW=0.9
momB=0.9
wc=0.0005
wball=0.00
schedW=jdexp[250,0,4]
schedB=dexp[10,0,2]

[conv2a]
epsW=0.01
epsB=0.02
momW=0.9
momB=0.9
wc=0.0005
wball=0.00
schedW=jdexp[250,0,4]
schedB=dexp[10,0,2]

[conv2b]
epsW=0.01
epsB=0.02
momW=0.9
momB=0.9
wc=0.0005
wball=0.00
schedW=jdexp[250,0,4]
schedB=dexp[10,0,2]

[conv3a]
epsW=0.01,0.01
epsB=0.02
momW=0.9,0.9
momB=0.9
wc=0.0005,0.0005
wball=0,0
schedW=jdexp[250,0,4]
schedB=dexp[10,0,2]

[conv3b]
epsW=0.01,0.01
epsB=0.02
momW=0.9,0.9
momB=0.9
wc=0.0005,0.0005
wball=0,0
schedW=jdexp[250,0,4]
schedB=dexp[10,0,2]

[conv4a]
epsW=0.01
epsB=0.02
momW=0.9
momB=0.9
wc=0.0005
wball=0
schedW=jdexp[250,0,4]
schedB=dexp[10,0,2]

[conv4b]
epsW=0.01
epsB=0.02
momW=0.9
momB=0.9
wc=0.0005
wball=0
schedW=jdexp[250,0,4]
schedB=dexp[10,0,2]

[conv5a]
epsW=0.01
epsB=0.02
momW=0.9
momB=0.9
wc=0.0005
wball=0
schedW=jdexp[250,0,4]
schedB=dexp[10,0,2]

[conv5b]
epsW=0.01
epsB=0.02
momW=0.9
momB=0.9
wc=0.0005
wball=0
schedW=jdexp[250,0,4]
schedB=dexp[10,0,2]

[fc2048a]
epsW=0.01,0.01
epsB=0.02
momW=0.9,0.9
momB=0.9
wc=0.0005,0.0005
wball=0,0
schedW=jdexp[250,0,4]
schedB=dexp[10,0,2]

[fc2048b]
epsW=0.01,0.01
epsB=0.02
momW=0.9,0.9
momB=0.9
wc=0.0005,0.0005
wball=0,0
schedW=jdexp[250,0,4]
schedB=dexp[10,0,2]

[fc2048ba]
epsW=0.01,0.01
epsB=0.02
momW=0.9,0.9
momB=0.9
wc=0.0005,0.0005
wball=0,0
schedW=jdexp[250,0,4]
schedB=dexp[10,0,2]

[fc2048bb]
epsW=0.01,0.01
epsB=0.02
momW=0.9,0.9
momB=0.9
wc=0.0005,0.0005
wball=0,0
schedW=jdexp[250,0,4]
schedB=dexp[10,0,2]

[fc1000]
epsW=0.01,0.01
epsB=0.02
momW=0.9,0.9
momB=0.9
wc=0.0005,0.0005
wball=0,0
schedW=jdexp[250,0,4]
schedB=dexp[10,0,2]

[logprob]
coeff=1
topk=5

[hs1a]
enable=true

[hs2a]
enable=true

[hs1b]
enable=true

[hs2b]
enable=true

[rnorm1a]
scale=0.0001
pow=0.75
minDiv=2

[rnorm1b]
scale=0.0001
pow=0.75
minDiv=2

[rnorm2a]
scale=0.0001
pow=0.75
minDiv=2

[rnorm2b]
scale=0.0001
pow=0.75
minDiv=2

[cnorm2a]
scale=0.001
pow=0.75

[cnorm2b]
scale=0.001
pow=0.75

# this is like #159 (so uses same def file: 153) but with jdexp
# this is also like 174, which uses jdexp, but 172 decayed learning rate by factor of 500 instead of 250
# so this one should be compared to 174 (which decays learning rate less) and 170 (which decays learning rate the same, but is not jumpy)
# on gpu
# initialized from 24 epochs of #169
# logs/layers-180.log
# /storage/tmp/ConvNet__2013-01-07_04.03.15.layers-180