Model: fish_4layer

simulations / paper

date: Thu, 29 May 2008 14:04:16
based on SVN revision: 406
original path:
/home/wilbert/simulations/paper/models/fish_4layer

parameters:

location servling
distributed True
computers (hamilton, hamilton, huxley, huxley, weismann, weismann, verhulst, verhulst, mendel, mendel, golgi, spemann, spemann, helmholtz, turing, monod, hebb, weinberg, weinberg, itb4, itb4, itb4, itb4, itb4, itb4, itb4, itb4)

layer_1 was trained with input fish_layer0_15_standard
/home/wilbert/simulations/paper/inputs/fish_layer0_15_standard/views
and was loaded from this model, from file
/home/wilbert/simulations/paper/models/fish_4layer/layers/layer_1.pckl

layer_2 will be trained with input fish_layer0_15_standard
/home/wilbert/simulations/paper/inputs/fish_layer0_15_standard/views

layer_3 will be trained with input fish_layer2_15_standard
/home/wilbert/simulations/paper/inputs/fish_layer2_15_standard/views

layer_4 will be trained with input fish_layer2_15_standard
/home/wilbert/simulations/paper/inputs/fish_layer2_15_standard/views

Model Structure


in-dim: 24336
in-dim: 24336
Rectangular2dSwitchboard
rec. field size (in channels): 10 x 10 = 100
# of rec. fields (output channels): 30 x 30 = 900
rec. field distances (in channels): (5, 5)
channel width: 1
out-dim: 90000
in-dim: 90000
ParallelCloneLayer

900 repetitions
in-dim: 100
NormalNoiseNode
noise level: 1e-06
noise offset: 0
out-dim: 100
in-dim: 100
ParallelSFANode
out-dim: 32
in-dim: 32
ParallelSFA2Node
expansion dim: 560
out-dim: 32
in-dim: 32
CutoffNode
lower bound: -4
upper bound: 4
out-dim: 32
out-dim: 28800
in-dim: 28800
Rectangular2dSwitchboard
rec. field size (in channels): 4 x 4 = 16
# of rec. fields (output channels): 14 x 14 = 196
rec. field distances (in channels): (2, 2)
channel width: 32
out-dim: 100352
in-dim: 100352
ParallelCloneLayer

196 repetitions
in-dim: 512
NormalNoiseNode
noise level: 1e-06
noise offset: 0
out-dim: 512
in-dim: 512
ParallelSFANode
out-dim: 32
in-dim: 32
ParallelSFA2Node
expansion dim: 560
out-dim: 32
in-dim: 32
CutoffNode
lower bound: -4
upper bound: 4
out-dim: 32
out-dim: 6272
in-dim: 6272
Rectangular2dSwitchboard
rec. field size (in channels): 4 x 4 = 16
# of rec. fields (output channels): 6 x 6 = 36
rec. field distances (in channels): (2, 2)
channel width: 32
out-dim: 18432
in-dim: 18432
ParallelCloneLayer

36 repetitions
in-dim: 512
NormalNoiseNode
noise level: 1e-06
noise offset: 0
out-dim: 512
in-dim: 512
ParallelSFANode
out-dim: 42
in-dim: 42
ParallelSFA2Node
expansion dim: 945
out-dim: 32
in-dim: 32
CutoffNode
lower bound: -4
upper bound: 4
out-dim: 32
out-dim: 1152
in-dim: 1152
Rectangular2dSwitchboard
rec. field size (in channels): 6 x 6 = 36
# of rec. fields (output channels): 1 x 1 = 1
rec. field distances (in channels): (6, 6)
channel width: 32
out-dim: 1152
in-dim: 1152
ParallelCloneLayer

1 repetitions
in-dim: 1152
NormalNoiseNode
noise level: 1e-06
noise offset: 0
out-dim: 1152
in-dim: 1152
ParallelSFANode
out-dim: 52
in-dim: 52
ParallelSFA2Node
expansion dim: 1430
out-dim: 512
in-dim: 512
CutoffNode
lower bound: -4
upper bound: 4
out-dim: 512
out-dim: 512
out-dim: 512


Test Data

test input fish_test_25_standard loaded from
/home/wilbert/simulations/paper/inputs/fish_test_25_standard/views

result written to
/home/wilbert/simulations/paper/models/fish_4layer/results/fish_test_25_standard.mat test input fish_layer0_15_standard loaded from
/home/wilbert/simulations/paper/inputs/fish_layer0_15_standard/views

result written to
/home/wilbert/simulations/paper/models/fish_4layer/results/fish_layer0_15_standard.mat

Optimal Stimuli

Extrema for a norm value of 64 per pixel.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
max
min