# 3.7.1. renom.utility.convergence package ¶

class  ConvergenceCheck  ( start_count=0 )

Super class of class ConvergenceCheckAbsoluteError and ConvergenceCheckRelativeError.

class  ConvergenceCheckAbsoluteError  ( start_count=0 )

“Check test loss and train loss whether they are convergent or not. The convergence test is based on value. The method is different between current value and previous value. This function returns Ture or False.If inputted data is convergence, the function returns True.

class  ConvergenceCheckRelativeError  ( start_count=0 )

Check test loss and train loss whether they are convergent or not. The convergence test is based on change rate. The method is different between current value and previous value. This function returns Ture or False.If inputted data is convergence, the function returns True.

 is_converged_relative_error  ( start_count , value , loss )

This function tests train loss data and test loss data whether they are convergent or not. The converfence test is decided that these loss function are convegent by change rate and the standard value is Parameter value.

 パラメータ: start_count ( int ) – Specified numbers of list value ( float ) – The determination value for convergence test loss ( list ) – Test data or test data boolean: Inputted data is convergence or not.
\begin{split}\left| \frac{X_{n-1}-X_{n}}{X_{n}} \right| < A\end{split}
X_{n-1}:Previous\ value
X_{n}:Current\ value
A:Standard\ value
>>> from renom.utility.convergence.convergence_check import is_converged_relative_error
>>> result = is_converged_relative_error(3,0.001,train)
>>> result
True
>>> result = is_converged_relative_error(3,0.001,test)
>>> result
False

 is_converged_absolute_error  ( start_count , value , loss )

This function tests train loss data and test loss data whether they are convergent or not. The converfence test is decided that these loss function are convergent by absolute error value and the standard value is Parameter value.

 パラメータ: start_count ( int ) – specified numbers of list. value ( float ) – the determination value for convergence test. loss ( list ) – test data or test data. boolean: inputted data is convergence or not.
\begin{split}|X_{n-1}-X_{n}| < A\end{split}
X_{n-1}: Previous\ value
X_{n}: current\ value
A:Standard\ value
>>> from renom.utility.convergence.convergence_check import is_converged_absolute_error
>>> result = is_converged_absolute_error(3,0.001,train)
>>> result
True
>>> result = is_converged_absolute_error(3,0.001,test)
>>> result
False