Weight Initialization¶
Weight initialization is a technique that deals with presetting of the weight parameters in the network prior to training in a manner that favors the ultimate convergence of the neural network.
Featured Initializers¶
One |
One (One) |
Zero |
Zero (Zero) |
HeNormal |
He Normal (HeNormal) |
HeUniform |
He Normal (HeNormal) |
GlorotNormal |
Glorot Normal (GlorotNormal) |
GlorotUniform |
Glorot Uniform (GlorotUniform) |
LeCunNormal |
LeCun Normal (LeCunNormal) |
LeCunUniform |
LeCun Uniform (LeCunUniform) |
RandomNormal |
Random Normal (RandomNormal) |
RandomUniform |
Random Uniform (RandomUniform) |
Function Descriptions¶
-
class
ztlearn.initializers.
GlorotNormal
[source]¶ Bases:
ztlearn.initializers.WeightInitializer
Glorot Normal (GlorotNormal)
GlorotNormal, more famously known as the Xavier initialization is based on the effort to try mantain the same variance of the gradients of the weights for all the layers. Glorot normal is an implementation based on Gaussian distribution
References
- [1] Understanding the difficulty of training deep feedforward neural networks
- [Xavier Glorot, 2010] http://proceedings.mlr.press/v9/glorot10a.html
- [PDF] http://proceedings.mlr.press/v9/glorot10a/glorot10a.pdf
- [2] Initialization Of Deep Feedfoward Networks
- [DeepGrid Article - Jefkine Kafunah] https://goo.gl/E2XrGe
-
init_name
¶
-
class
ztlearn.initializers.
GlorotUniform
[source]¶ Bases:
ztlearn.initializers.WeightInitializer
Glorot Uniform (GlorotUniform)
GlorotUniform, more famously known as the Xavier initialization is based on the effort to try mantain the same variance of the gradients of the weights for all the layers. Glorot uniform is an implementation based on Uniform distribution
References
- [1] Understanding the difficulty of training deep feedforward neural networks
- [Xavier Glorot, 2010] http://proceedings.mlr.press/v9/glorot10a.html
- [PDF] http://proceedings.mlr.press/v9/glorot10a/glorot10a.pdf
- [2] Initialization Of Deep Feedfoward Networks
- [DeepGrid Article - Jefkine Kafunah] https://goo.gl/E2XrGe
-
init_name
¶
-
class
ztlearn.initializers.
HeNormal
[source]¶ Bases:
ztlearn.initializers.WeightInitializer
He Normal (HeNormal)
HeNormal is a robust initialization method that particularly considers the rectifier nonlinearities. He normal is an implementation based on Gaussian distribution
References
- [1] Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification
- [Kaiming He, 2015] https://arxiv.org/abs/1502.01852
- [PDF] https://arxiv.org/pdf/1502.01852.pdf
- [2] Initialization Of Deep Networks Case of Rectifiers
- [DeepGrid Article - Jefkine Kafunah] https://goo.gl/TBNw5t
-
init_name
¶
-
class
ztlearn.initializers.
HeUniform
[source]¶ Bases:
ztlearn.initializers.WeightInitializer
He Normal (HeNormal)
HeNormal is a robust initialization method that particularly considers the rectifier nonlinearities. He uniform is an implementation based on Uniform distribution
References
- [1] Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification
- [Kaiming He, 2015] https://arxiv.org/abs/1502.01852
- [PDF] https://arxiv.org/pdf/1502.01852.pdf
- [2] Initialization Of Deep Networks Case of Rectifiers
- [DeepGrid Article - Jefkine Kafunah] https://goo.gl/TBNw5t
-
init_name
¶
-
class
ztlearn.initializers.
Identity
[source]¶ Bases:
ztlearn.initializers.WeightInitializer
Identity (Identity)
Identity is an implementation of weight initialization that returns an identity matrix of size shape
-
init_name
¶
-
-
class
ztlearn.initializers.
LeCunNormal
[source]¶ Bases:
ztlearn.initializers.WeightInitializer
LeCun Normal (LeCunNormal)
Weights should be randomly chosen but in such a way that the sigmoid is primarily activated in its linear region. LeCun uniform is an implementation based on Gaussian distribution
References
- [1] Efficient Backprop
- [LeCun, 1998][PDF] http://yann.lecun.com/exdb/publis/pdf/lecun-98b.pdf
-
init_name
¶
-
class
ztlearn.initializers.
LeCunUniform
[source]¶ Bases:
ztlearn.initializers.WeightInitializer
LeCun Uniform (LeCunUniform)
Weights should be randomly chosen but in such a way that the sigmoid is primarily activated in its linear region. LeCun uniform is an implementation based on Uniform distribution
References
- [1] Efficient Backprop
- [LeCun, 1998][PDF] http://yann.lecun.com/exdb/publis/pdf/lecun-98b.pdf
-
init_name
¶
-
class
ztlearn.initializers.
One
[source]¶ Bases:
ztlearn.initializers.WeightInitializer
One (One)
One is an implementation of weight initialization that returns all ones
-
init_name
¶
-
-
class
ztlearn.initializers.
RandomNormal
[source]¶ Bases:
ztlearn.initializers.WeightInitializer
Random Normal (RandomNormal)
Random uniform, an implementation of weight initialization based on Gaussian distribution
-
init_name
¶
-
-
class
ztlearn.initializers.
RandomUniform
[source]¶ Bases:
ztlearn.initializers.WeightInitializer
Random Uniform (RandomUniform)
Random uniform, an implementation of weight initialization based on Uniform distribution
-
init_name
¶
-