next up previous contents index
Next: Complexity of SCG Up: Scaled Conjugate Gradient Previous: Main features of

Parameters of SCG

As and are computed from their respective values at step k-1, SCG has two parameters, namely the initial values and . Their values are not critical but should respect the conditions and . Empirically Møller has shown that bigger values of can lead to a slower convergence.

The third parameter is the usual quantity (cf. standard backpropagation).

In SNNS, it is usually the responsibility of the user to determine when the learning process should stop. Unfortunately, the adaptation mechanism sometimes assigns too large values to when no more progress is possible. In order to avoid floating-point exceptions, we have added a termination criteriongif to SCG. The criterion is taken from the CGMs presented in [P88]: stop when

is a small number used to rectify the special case of converging to a function value of exactly zero. It is set to . is a tolerance depending of the floating-point precision of your machine, and it should be set to , which is usually equal to (simple precision) or to (double precision).

To summarize, there are four non-critical parameters:

  1. . Should satisfy . If 0, will be set to ;
  2. . Should satisfy . If 0, will be set to ;
  3. . See standard backpropagation. Can be set to 0 if you don't know what to do with it;
  4. . Depends on the floating-point precision. Should be set to (simple precision) or to (double precision). If 0, will be set to .

Note: SCG is a batch learning method, so shuffling the patterns has no effect.



Niels.Mache@informatik.uni-stuttgart.de
Tue Nov 28 10:30:44 MET 1995