def res_net_pyramidal_model_d6_w32_k2_dr05(
features,
targets,
mode,
optimizer_type='SGD',
learning_rate=0.001):
return model.res_net_pyramidal_model(
features=features,
targets=targets,
mode=mode,
num_classes=2,
num_blocks=int(6/3),
multi_k=2,
keep_prob=0.5,
optimizer_type=optimizer_type,
learning_rate=learning_rate,
groups=[16, 16, 32, 32],
scope="rnp_d6_w32_k2_dr05")
def res_net_pyramidal_model_d21_w256_k4_dr05(
features,
targets,
mode,
optimizer_type='SGD',
learning_rate=0.001):
return model.res_net_pyramidal_model(
features=features,
targets=targets,
mode=mode,
num_classes=2,
num_blocks=int(21/3),
multi_k=4,
keep_prob=0.5,
optimizer_type=optimizer_type,
learning_rate=learning_rate,
groups=[16, 64, 128, 256],
scope="rnp_d21_w128_k4_dr05")
def res_net_pyramidal_model_d110_w350(
features,
targets,
mode,
optimizer_type='SGD',
learning_rate=0.001):
""" Deep Pyramidal Residual Networks
From https://arxiv.org/abs/1610.02915
"""
return model.res_net_pyramidal_model(
features=features,
targets=targets,
mode=mode,
num_classes=2,
num_blocks=36,
optimizer_type=optimizer_type,
learning_rate=learning_rate,
groups=[16, 150, 250, 350],
scope="rnp_d110_w350")
def res_net_pyramidal_model_d12_w256_k2_dr05_ds(
features,
targets,
mode,
optimizer_type='SGD',
learning_rate=0.001):
return model.res_net_pyramidal_model(
features=features,
targets=targets,
mode=mode,
num_classes=2,
num_blocks=int(12/3),
multi_k=2,
keep_prob=0.5,
optimizer_type=optimizer_type,
learning_rate=learning_rate,
groups=[16, 32, 128, 256],
is_double_size=True,
scope="rnp_d12_w256_k2_dr05_ds")
