def __init__(self, opt={}):
self.out_depth = opt['in_depth']
self.out_sx = opt['in_sx']
self.out_sy = opt['in_sy']
self.num_inputs = opt['in_sx'] * opt['in_sy'] * opt['in_depth']
self.layer_type = 'add'
self.skip = getopt(opt, 'skip', 0) # skip n activations in input
self.delta = getopt(opt, 'delta', [0] * self.num_inputs)
self.num_neurons = getopt(opt, 'num_neurons', self.num_inputs)
def __init__(self, opt={}):
self.out_depth = opt['in_depth']
self.out_sx = opt['in_sx']
self.out_sy = opt['in_sy']
self.num_inputs = opt['in_sx'] * opt['in_sy'] * opt['in_depth']
self.layer_type = 'add'
self.skip = getopt(opt, 'skip', 0) # skip n activations in input
self.delta = getopt(opt, 'delta', [0] * self.num_inputs)
self.num_neurons = getopt(opt, 'num_neurons', self.num_inputs)
def __init__(self, opt={}):
self.out_depth = opt['num_neurons']
self.l1_decay_mul = getopt(opt, 'l1_decay_mul', 0.0)
self.l2_decay_mul = getopt(opt, 'l2_decay_mul', 1.0)
self.num_inputs = opt['in_sx'] * opt['in_sy'] * opt['in_depth']
self.out_sx = 1
self.out_sy = 1
self.layer_type = 'fc'
bias = getopt(opt, 'bias_pref', 0.0)
self.filters = [ Vol(1, 1, self.num_inputs) for i in xrange(self.out_depth) ]
self.biases = Vol(1, 1, self.out_depth, bias)
def __init__(self, opt={}):
self.out_depth = opt['num_neurons']
self.l1_decay_mul = getopt(opt, 'l1_decay_mul', 0.0)
self.l2_decay_mul = getopt(opt, 'l2_decay_mul', 1.0)
self.num_inputs = opt['in_sx'] * opt['in_sy'] * opt['in_depth']
self.out_sx = 1
self.out_sy = 1
self.layer_type = 'sim'
bias = getopt(opt, 'bias_pref', 0.0)
self.filters = [ Vol(1, 1, self.num_inputs) for i in xrange(self.out_depth) ]
self.biases = Vol(1, 1, self.out_depth, bias)
def __init__(self, opt={}):
self.group_size = getopt(opt, 'group_size', 2)
self.out_sx = opt['in_sx']
self.out_sy = opt['in_sy']
self.out_depth = opt['in_depth'] / self.group_size
self.layer_type = 'maxout'
self.switches = zeros(self.out_sx * self.out_sy * self.out_depth)
def __init__(self, opt={}):
self.out_sx = opt['in_sx']
self.out_sy = opt['in_sy']
self.out_depth = opt['in_depth']
self.layer_type = 'dropout'
self.drop_prob = getopt(opt, 'drop_prob', 0.5)
self.dropped = zeros(self.out_sx * self.out_sy * self.out_depth)
def __init__(self, opt={}):
self.group_size = getopt(opt, 'group_size', 2)
self.out_sx = opt['in_sx']
self.out_sy = opt['in_sy']
self.out_depth = opt['in_depth'] / self.group_size
self.layer_type = 'maxout'
self.switches = zeros(self.out_sx * self.out_sy * self.out_depth)
def __init__(self, opt={}):
self.out_sx = opt['in_sx']
self.out_sy = opt['in_sy']
self.out_depth = opt['in_depth']
self.zeta = getopt(opt, 'zeta', 1.0)
if self.zeta == 0.0:
print 'WARNING: zeta cannot equal 0'
self.layer_type = 'mex'
def __init__(self, opt={}):
self.sx = opt['sx'] # filter size
self.in_depth = opt['in_depth']
self.in_sx = opt['in_sx']
self.in_sy = opt['in_sy']
# optional
self.sy = getopt(opt, 'sy', self.sx)
self.stride = getopt(opt, 'stride', 2)
self.pad = getopt(opt, 'pad', 0) # padding to borders of input volume
self.out_depth = self.in_depth
self.out_sx = int(floor((self.in_sx - self.sx + 2 * self.pad) / self.stride + 1))
self.out_sy = int(floor((self.in_sy - self.sy + 2 * self.pad) / self.stride + 1))
self.layer_type = 'pool'
# Store switches for x,y coordinates for where the max comes from, for each output neuron
switch_size = self.out_sx * self.out_sy * self.out_depth
self.switch_x = zeros(switch_size)
self.switch_y = zeros(switch_size)
def __init__(self, opt={}):
self.out_depth = opt['filters']
self.sx = opt['sx'] # filter size: should be odd if possible
self.in_depth = opt['in_depth']
self.in_sx = opt['in_sx']
self.in_sy = opt['in_sy']
# optional
self.sy = getopt(opt, 'sy', self.sx)
self.stride = getopt(opt, 'stride', 1) # stride at which we apply filters to input volume
self.pad = getopt(opt, 'pad', 0) # padding to borders of input volume
self.l1_decay_mul = getopt(opt, 'l1_decay_mul', 0.0)
self.l2_decay_mul = getopt(opt, 'l2_decay_mul', 1.0)
"""
Note we are doing floor, so if the strided convolution of the filter doesnt fit into the input
volume exactly, the output volume will be trimmed and not contain the (incomplete) computed
final application.
"""
self.out_sx = int(floor((self.in_sx - self.sx + 2 * self.pad) / self.stride + 1))
self.out_sy = int(floor((self.in_sy - self.sy + 2 * self.pad) / self.stride + 1))
self.layer_type = 'conv'
bias = getopt(opt, 'bias_pref', 0.0)
self.filters = [ Vol(self.sx, self.sy, self.in_depth) for i in xrange(self.out_depth) ]
self.biases = Vol(1, 1, self.out_depth, bias)
def __init__(self, opt={}):
self.out_depth = opt['filters']
self.sx = opt['sx'] # filter size: should be odd if possible
self.in_depth = opt['in_depth']
self.in_sx = opt['in_sx']
self.in_sy = opt['in_sy']
# optional
self.sy = getopt(opt, 'sy', self.sx)
self.stride = getopt(
opt, 'stride',
1) # stride at which we apply filters to input volume
self.pad = getopt(opt, 'pad', 0) # padding to borders of input volume
self.l1_decay_mul = getopt(opt, 'l1_decay_mul', 0.0)
self.l2_decay_mul = getopt(opt, 'l2_decay_mul', 1.0)
"""
Note we are doing floor, so if the strided convolution of the filter doesnt fit into the input
volume exactly, the output volume will be trimmed and not contain the (incomplete) computed
final application.
"""
self.out_sx = int(
floor((self.in_sx - self.sx + 2 * self.pad) / self.stride + 1))
self.out_sy = int(
floor((self.in_sy - self.sy + 2 * self.pad) / self.stride + 1))
self.layer_type = 'conv'
bias = getopt(opt, 'bias_pref', 0.0)
self.filters = [
Vol(self.sx, self.sx, self.in_depth)
for i in xrange(self.out_depth)
]
self.biases = Vol(1, 1, self.out_depth, bias)
