NAN error on small weights. #41

Open
samWighton opened this Issue Nov 8, 2016 · 0 comments

Projects

None yet

1 participant

@samWighton

Using default settings except:
Problem type = regression
Zero hidden layers
3 features

When training starts, the third feature will quickly move to a very tiny value.
Shortly after, the training stops:
screen shot 2016-11-08 at 11 45 56 am

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment