Skip to content

Latest commit

 

History

History
20 lines (14 loc) · 638 Bytes

log.md

File metadata and controls

20 lines (14 loc) · 638 Bytes
Raw

AOSOLogitBoost

The LogitBoost implementation described in "Sun P., Reid, M.D., Zhou J. AOSO-LogitBoost: Adaptive One-Vs-One LogitBoost for Multi-Class Problems, ICML 2012"

pAOSOLogitBoost

Added on 2013-7.

Parallel version (multi-threaded) of AOSOLogitBoost.

pAOSOLogitBoostV2

Added on 2013-10-20.

Improvements to pAOSOLogitBoost:

  • Speedup (see matlab/run_script/run_pAOSOLogitBoostV2_xxx.m) of training by 1) subsampling instances (Friedman's weight trimming) and 2) subsampling features (uniformly sampling at each tree node)
  • bugs fixed (crashing due to null leaf)
  • redundant computations eliminated