link logistic error rates

[nikete]

When using the logistic we get crazy high errors reported under --binary loss

for the plain linear model with logistic link and loss at 65536 :
0.378387

while withoutany logistic:
0.013412

while with only the loss logistic
0.016068

are we fitting that badly or is it not the bianry loss that gets reported?

[ekg]

We might need to take this to vw chat or mailing list.

On Tue, Apr 12, 2016, 15:56 Nicolás Della Penna notifications@github.com
wrote:

When using the logistic we get crazy high errors reported under --binary
loss

for the plain linear model with logistic link and loss at 65536 :
0.378387

while withoutany logistic:
0.013412

while with only the loss logistic
0.016068

are we fitting that badly or is it not the bianry loss that gets reported?

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub
#14

[nikete]

it is a reporting bug from vw