Udacity: Deep Learning: Lesson 3: Training Neural Networks

[mbhushan]

Momentum:

Random Restarts:

Learning Rate characteristics:

Learning Rate:

Backpropogation with activation functions such as Relu, tanh - addresses vanishing gradient:

Rectified Linear Unit (Relu):

Hyperbolic tangent function:

Backpropogation:

Local Minima:

Regularization:

Activation Functions Error:

Model Complexity Graph:

Early Stopping:

Classification: Overfitting

Type of Errors: