def __init__(self, learning_rate):
n_features = eval('*'.join(str(e) for e in DB.feature_shape))
n_classes = DB.n_classes
self.lr = nn.parameter(learning_rate)
# Step 3.1: set up learnable parameters/weights
#######################################################################
"*** YOUR CODE HERE ***"
#######################################################################
self.param = {
"w1": nn.parameter(np.random.uniform(-0.01, 0.01, size=[n_features,neurons])),
"w2": nn.parameter(np.random.uniform(-0.01, 0.01, size=[neurons, n_classes])),
"b1": nn.parameter(np.random.uniform(-0.01, 0.01, size=[neurons])),
"b2": nn.parameter(np.random.uniform(-0.01, 0.01, size=[n_classes])),
}
def __init__(self, learning_rate):
n_features = eval('*'.join(str(e) for e in DB.feature_shape))
n_classes = DB.n_classes
self.lr = nn.parameter(learning_rate)
# Step 3.1: set up learnable parameters/weights. w's should be drawn
# from a zero-mean normal distribution. b's should be set to 0
#######################################################################
"*** YOUR CODE HERE ***"
#######################################################################
self.param = {
'w1':
nn.parameter.from_numpy(
numpy.random.normal(0, 0.01, [n_features, 256])),
'w2':
nn.parameter.from_numpy(numpy.random.normal(0, 0.01, [256, 256])),
'w3':
nn.parameter.from_numpy(numpy.random.normal(0, 0.01, [256, 256])),
'w4':
nn.parameter.from_numpy(numpy.random.normal(0, 0.01, [256, 256])),
'w5':
nn.parameter.from_numpy(
numpy.random.normal(0, 0.01, [256, n_classes])),
'b1':
nn.parameter.zeros(256),
'b2':
nn.parameter.zeros(256),
'b3':
nn.parameter.zeros(256),
'b4':
nn.parameter.zeros(256),
'b5':
nn.parameter.zeros(n_classes)
}
def __init__(self, learning_rate):
n_features = eval('*'.join(str(e) for e in DB.feature_shape))
n_classes = DB.n_classes
self.lr = nn.parameter(learning_rate)
# Step 3.1: set up learnable parameters/weights. w's should be drawn
# from a zero-mean normal distribution. b's should be set to 0
#######################################################################
"*** YOUR CODE HERE ***"
#######################################################################
self.param = {
# "name": nn.parameter(blah, blah)
"w": nn.parameter(np.random.uniform(-0.1, 0.1, [n_features, 512])),
"b": nn.parameter.zeros([512]),
"w1": nn.parameter(np.random.uniform(-0.05, 0.05, [512, 256])),
"b1": nn.parameter.zeros([256]),
"w2": nn.parameter(np.random.uniform(-0.01, 0.01,
[256, n_classes])),
"b2": nn.parameter.zeros([n_classes]),
}
def __init__(self, learning_rate):
n_features = eval('*'.join(str(e) for e in DB.feature_shape))
n_classes = DB.n_classes
self.lr = nn.parameter(learning_rate)
# Step 3.1: set up learnable parameters/weights
#######################################################################
"*** YOUR CODE HERE ***"
#######################################################################
self.param = {
"w1":
nn.parameter(
np.add(np.zeros((n_features, intermediate_val_1)),
np.random.rand(n_features, intermediate_val_1) - 0.5)),
"w2":
nn.parameter(
np.add(
np.zeros((intermediate_val_1, intermediate_val_2)),
np.random.rand(intermediate_val_1, intermediate_val_2) -
0.5)),
"w3":
nn.parameter(
np.add(
np.zeros((intermediate_val_2, intermediate_val_3)),
np.random.rand(intermediate_val_2, intermediate_val_3) -
0.5)),
"w4":
nn.parameter(
np.add(np.zeros((intermediate_val_3, n_classes)),
np.random.rand(intermediate_val_3, n_classes) - 0.5)),
"b1":
nn.parameter.zeros(intermediate_val_1),
"b2":
nn.parameter.zeros(intermediate_val_2),
"b3":
nn.parameter.zeros(intermediate_val_3),
"b4":
nn.parameter.zeros(n_classes),
}
def __init__(self, learning_rate):
n_features = eval('*'.join(str(e) for e in DB.feature_shape))
n_classes = DB.n_classes
self.lr = nn.parameter(learning_rate)
# Step 3.1: set up learnable parameters/weights
#######################################################################
"*** YOUR CODE HERE ***"
#######################################################################
self.param = {
#Weight1
"w1":
nn.parameter(
np.asmatrix(
np.random.rand(n_features, hidden_layer_level_1_size) -
0.50)),
#Weight2
"w2":
nn.parameter(
np.asmatrix(
np.random.rand(hidden_layer_level_1_size,
hidden_layer_level_2_size) - 0.5)),
#Weight3
"w3":
nn.parameter(
np.asmatrix(
np.random.rand(hidden_layer_level_2_size, n_classes) -
0.50)),
#Bias1
"b1":
nn.parameter.zeros(hidden_layer_level_1_size),
#Bias2
"b2":
nn.parameter.zeros(hidden_layer_level_2_size),
#Bias3
"b3":
nn.parameter.zeros(n_classes),
}