About this document
Majorization Techniques in Optimization by Vikas Kumar is a document available to read on EtoBox.
The document discusses majorization techniques for optimization problems that lack closed-form solutions. It presents a tighter quadratic bound for partition functions of log-linear models that can be used within majorization methods. The bound is initialized and then updated iteratively based on the features and weights of each configuration. Finally, it notes the bound leads to faster majorization methods and can find better local maxima when applied to problems involving graphical models, high-dimensiona
- Author
- Vikas Kumar
- Language
- EN