DBN Class
Deep Belief Nets
Constructor
DBN
()
Methods
predict
(
-
x
Predict label with training data
Parameters:
-
x
Matrixinput data
pretrain
(
-
lr
-
k
-
epochs
Training hidden layers with unsupervised learning
Parameters:
-
lr
Floatlearning rate
-
k
Intthe number of phase
-
epochs
Intthe number of times of running gradient decent
train
(
-
lr
-
epochs
Training logistics regression algorithm which is on output layer
Parameters:
-
lr
Floatlearning rate
-
epochs
Intthe number of times of running gradient decent
Properties
nLayers
Int
number of hidden layers
sigmoidLayers
Array
hidden layers which activations are sigmoid function
x
Matrix
input data. This type is defined in sylvester library
y
Matrix
label data. This type is defined in sylvester library