RM4.6 vs RM5 differences in OptimizeParameters Operator

Axel

Hi,

I noticed a strange difference in the Optimize Parameters operator (GridParameterOptimization) between RM4.6 and RM5.
I'm varying a parameter from 1e-5 to 1 in 9 steps using a logarithmic scale which gives me in 4.6 the values:
0.00001
0.08007
0.16654
0.25993
0.36080
0.46974
0.58741
0.71449
0.85175
1

while in RM5 I get:
0.00001
0.00004
0.00013
0.00046
0.00167
0.00599
0.02154
0.07743
0.27826
1

Why is this ?
It makes it very difficult to compare the results between the two versions.

Many thanks,

Axel

land

Hi Axel,
actually the formula was changed to better fit the needs especially for C parameter of SVMs. With the next update I there will be an additional scale called "logarithmic (legacy)" which will be compatible with the old legacy mode.

Greetings,
  Sebastian