keraflow.constraints.MaxNorm Class Reference

Constrain the weights along an axis pattern to have unit norm. More...

Inheritance diagram for keraflow.constraints.MaxNorm:

Public Member Functions
def	__init__

Detailed Description

Constrain the weights along an axis pattern to have unit norm.

Constructor & Destructor Documentation

def keraflow.constraints.MaxNorm.__init__	(		self,
			m = 2,
			axis = 0
	)

Parameters

m

float. The target max norm.

axis

int/list of int, axis pattern along which to calculate weight norms. Examples: In Dense layer, W has has shape (input_dim, output_dim), set axis to 0 to constrain each row vector of size (input_dim,). In a Convolution2D, W has shape (input_depth, output_depth, kernel_rows, kernel_cols), set axis to [0,2,3] to constrain the weights of each kernel tensor of size (input_depth, kernel_rows, kernel_cols).