Timezone: »

Logarithmic Time One-Against-Some
Hal Daumé · Nikos Karampatziakis · John Langford · Paul Mineiro

Sun Aug 06 09:24 PM -- 09:42 PM (PDT) @ C4.6 & C4.7

We create a new online reduction of multiclass classification to binary classification for which training and prediction time scale logarithmically with the number of classes. We show that several simple techniques give rise to an algorithm which is superior to previous logarithmic time classification approaches while competing with one-against-all in space. The core construction is based on using a tree to select a small subset of labels with high recall, which are then scored using a one-against-some structure with high precision.

Author Information

Hal Daumé (University of Maryland)
Nikos Karampatziakis (Microsoft)
John Langford (Microsoft Research)
Paul Mineiro (Microsoft)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors