 
Summary: A Note on Platt's Probabilistic Outputs for Support
Vector Machines
HsuanTien Lin (htlin@ntu.edu.tw)
ChihJen Lin (cjlin@csie.ntu.edu.tw)
Department of Computer Science and Information Engineering,
National Taiwan University, Taipei 106, Taiwan
Ruby C. Weng (chweng@nccu.edu.tw)
Department of Statistics,
National Chengchi University, Taipei 116, Taiwan
Abstract. Platt's probabilistic outputs for Support Vector Machines (Platt, 2000)
has been popular for applications that require posterior class probabilities. In this
note, we propose an improved algorithm that theoretically converges and avoids
numerical difficulties. A simple and readytouse pseudo code is included.
Keywords: Support Vector Machine, Posterior Probability
1. Introduction
Given training examples xi Rn, i = 1, . . . , l, labeled by yi {+1, 1},
the binary Support Vector Machine (SVM) computes a decision func
tion f(x) such that sign(f(x)) can be used to predict the label of any
test example x.
Instead of predicting the label, many applications require a posterior
