Beta. Content is under active construction and has not been peer-reviewed. Report errors on
GitHub
.
Disclaimer
Theorem
Path
Curriculum
Paths
Demos
Diagnostic
Search
Quiz Hub
/
Algorithmic Stability
Algorithmic Stability
1 questions
Difficulty 7-7
View topic
Advanced
0 / 1
1 advanced
Adapts to your performance
1 / 1
advanced (7/10)
conceptual
A
λ
-regularized ERM algorithm is
β
-uniformly stable with
β
=
O
(
1/
(
nλ
))
. What does this imply about the relationship between regularization strength and generalization?
Hide and think first
A.
Setting
λ
=
1/
n
always gives the optimal generalization rate, regardless of the problem
B.
The result only applies to convex losses, so it tells us essentially nothing about neural networks in practice
C.
Stronger regularization gives smaller
β
and tighter generalization but also increases bias
D.
Generalization is guaranteed for any
λ
>
0
, independently of the sample size
n
used to train
Submit Answer