Varying lambda
9 views
Skip to first unread message
sajit
Nov 12, 2011, 11:19:23 AM11/12/11
to NYC Machine Learning Review
It wasnt very clear from the videos (Regularization and bias
variance).. I wanted to confirm that as we change lambda , we are not
changing the degree of the hypothesis function.
He uses theta(1), theta(2) ,.theta(12) to represent theta for various
values of lambda. These were obtained by using the exactly same
hypothesis but different values of lambda - correct?
variance).. I wanted to confirm that as we change lambda , we are not
changing the degree of the hypothesis function.
He uses theta(1), theta(2) ,.theta(12) to represent theta for various
values of lambda. These were obtained by using the exactly same
hypothesis but different values of lambda - correct?
Paul Dix
Nov 12, 2011, 8:24:44 PM11/12/11
to nyc-machine-l...@googlegroups.com
That's right, changing lambda has no effect on the degree. It adjusts
how small or large the values of hypothesis function (the
coefficients) will be.
Small lambda = high variance = larger hypothesis values
Large lambda = high bias = smaller hypothesis values
how small or large the values of hypothesis function (the
coefficients) will be.
Small lambda = high variance = larger hypothesis values
Large lambda = high bias = smaller hypothesis values
Paul
Reply all
Reply to author
Forward
0 new messages