Repository logo
Log in(current)
Repository logoMIT Open ScholarshipDSpace@MIT
  1. Home
  2. Computer Science and Artificial Intelligence Lab (CSAIL)
  3. Artificial Intelligence Lab Publications
  4. AI Memos (1959 - 2004)
  5. Bagging Regularizes

Bagging Regularizes

Thumbnail Image
Download
Name

AIM-2002-003.pdf

Size

278.96 KB

Format

Adobe PDF

Checksum (MD5)

56ba19eb0f63dedb877b975d648206e4

Thumbnail Image
Download
Name

AIM-2002-003.ps

Size

885.08 KB

Format

PostScript

Checksum (MD5)

af596544af29954edaa6d7a05b986d63

Author(s)
Poggio, Tomaso
•
Rifkin, Ryan
•
Mukherjee, Sayan
•
Rakhlin, Alex
Date Issued
March 1, 2002
Series/Report no.
AIM-2002-003
CBCL-214
Abstract
Intuitively, we expect that averaging --- or bagging --- different regressors with low correlation should smooth their behavior and be somewhat similar to regularization. In this note we make this intuition precise. Using an almost classical definition of stability, we prove that a certain form of averaging provides generalization bounds with a rate of convergence of the same order as Tikhonov regularization --- similar to fashionable RKHS-based learning algorithms.
Subjects
AI
Bagging
stability
regularization
Persistent DSpace Link
http://hdl.handle.net/1721.1/7268
Repository logo
PrivacyPermissionsAccessibilityContact us
Repository logo
Notify us about copyright concerns.