MathUtils::Tanh test failing
I get this error:
Value of: tanh((x+_x0)/_lambda)-tanh((x-_x0)/_lambda)
Expected: 2.*_d_tanh.GetDoubleTanh(x, 0)
Which is: 0.67945553896577615
[ FAILED ] TanhTest.GetDoubleTanhTest (4 ms)
In a ton of places. Can you look?
It's not on all systems though.
Compares numerical derivative vs analytical derivative - so sensitive to machine precision... probably the issue...
Actually slightly different issue, but floating point precision problem nonetheless...
fixed in rogers dev, running hudson now...
Pushed fix to the trunk - but I can't reproduce it on my machine, so can only hope it is fixed...
ah right, you pushed it (I was trying to merge it and got really confused because the commit was there, but the timestamp was wrong). :)
I'll start the testing on the wide range of machines if they aren't already running
- Status changed from Open to Closed
- % Done changed from 0 to 100
Also available in: Atom