caffe2.python.regularizer.LogBarrier Class Reference

Inheritance diagram for caffe2.python.regularizer.LogBarrier:

Public Member Functions
def	__init__ (self, reg_lambda, discount_policy="inv", discount_options=None)
Public Member Functions inherited from caffe2.python.regularizer.Regularizer
def	__init__ (self)
def	__call__ (self, net, param_init_net, param, grad=None, by=None)

Public Attributes
	reg_lambda
	discount_policy
	discount_options
Public Attributes inherited from caffe2.python.regularizer.Regularizer
	kEpsilon

Detailed Description

Wright, S., & Nocedal, J. (1999). Numerical optimization. Springer Science,
35(67-68), 7. Chapter 19

Definition at line 140 of file regularizer.py.

Constructor & Destructor Documentation

def caffe2.python.regularizer.LogBarrier.__init__	(		self,
			reg_lambda,
			discount_policy = "inv",
			discount_options = None
	)

discount is a positive weight that is decreasing, and here it is implemented
similar to the learning rate. It is specified by a learning rate policy and
corresponding options

Definition at line 146 of file regularizer.py.

---

The documentation for this class was generated from the following file:

• caffe2/python/regularizer.py