171 search results - page 12 / 35 » A Boosting Approach to Multiple Instance Learning |
IJCAI
2007
13 years 9 months ago
2007
The training of support vector machines (SVM) involves a quadratic programming problem, which is often optimized by a complicated numerical solver. In this paper, we propose a muc...
ICDM
2009
IEEE
14 years 2 months ago
2009
IEEE
—One common approach to active learning is to iteratively train a single classifier by choosing data points based on its uncertainty, but it is nontrivial to design uncertainty ...
IJCAI
2003
13 years 9 months ago
2003
Ensemble methods like bagging and boosting that combine the decisions of multiple hypotheses are some of the strongest existing machine learning methods. The diversity of the memb...
ML
2012
ACM
12 years 3 months ago
2012
ACM
Dependency networks approximate a joint probability distribution over multiple random variables as a product of conditional distributions. Relational Dependency Networks (RDNs) are...
COLING
2010
13 years 2 months ago
2010
This paper presents a new approach to improving relation extraction based on minimally supervised learning. By adding some limited closed-world knowledge for confidence estimation...