keras之权重初始化方式

model.add(Dense(64, kernel_initializer=initializers.random_normal(stddev=0.01)))
 
# also works; will use the default parameters.
model.add(Dense(64, kernel_initializer='random_normal'))

keras.initializers.Zeros()#全0
keras.initializers.Ones()#全1
keras.initializers.RandomNormal(mean=0.0, stddev=0.05, seed=None))#指定均值和方差的正态分布初始化
keras.initializers.RandomUniform(minval=-0.05, maxval=0.05, seed=None)#指定下边界和上边界的均匀分布初始化
keras.initializers.TruncatedNormal(mean=0.0, stddev=0.05, seed=None)#截尾高斯分布初始化，位于均值两个标准差以外的数据将会被丢弃并重新生成，形成截尾分布

def my_init(shape, dtype=None):
 return K.random_normal(shape, dtype=dtype)
 
model.add(Dense(64, init=my_init))

from keras.models import Sequential
 
model =Sequential()
model.add(12,input_dim=8,kernel_initializer='random_uniform')