NIPS 2010 - More Highlights

This post is about all the other papers on NIPS that I found particularly interesting but don't have the time to write a lot about.
There is a similar post at Yaroslav Bulato's blog.
  • Multiple Kernel Learning and the SMO Algorithm by Vishwanathan, Sun, Ampornput and Varma (pdf) . Code here.
    Efficient training of p-norm MKL using Sequential Minimal Optimization.
  • Kernel Descriptors for Visual Recognition by
    Liefeng Bo, Xiaofeng Ren, Dieter Fox (pdf)
    A general setting to design image patch descriptors using kernels. The proposed kernel is demonstrated to outperform SIFT.
  • A Theory of Multiclass Boosting  by Indraneel Mukherjee, Robert Schapire (pdf)
    Title says it all.
  • Deep Coding Network  by Yuanqing Lin, Zhang Tong, Shenghuo Zhu, Kai Yu (pdf)
    This is a continuation of the work on Linear Coordinate Coding (pdf) which won the image net callenge.
  • Tree-Structured Stick Breaking for Hierarchical Data  by Ryan Adams, Zoubin Ghahramani, Michael Jordan (pdf)
    This is work on hierarchical topic models using trees. I must confess I have not ultimately understood this work but I find it highly interesting. In contrast to nested Chinese Restaurant Processes(pdf ), samples from the distribution can also be inner nodes.
  • Segmentation as Maximum-Weight Independent Set  by William Brendel, Sinisa Todorovic (pdf)
    This work on image segmentation formulates consistent segmentations using graphs and introduces a new optimization algorithm to approximately find independent sets.
    This work is quite similar to a tech report from Adrian Ion (pdf) but focuses more on optimization than on the potentials for the graph.
  • Simultaneous Object Detection and Ranking with Weak Supervision  by Matthew Blaschko, Andrea Vedaldi, Andrew Zisserman (pdf)
    A multiple instance approach to finding objects in weakly labeled images using something like a "latent SVR".
  • Convex Multiple-Instance Learning by Estimating Likelihood Ratio  by Fuxin Li, Cristian Sminchisescu (pdf)
    Multiple instance learning using a novel convex formulation. This is a promising direction and hopefully will be applied soon to object segmentation using "Object Recognition as Ranking Holistic Figure-Ground Hypotheses" (pdf) by the same group.


Popular posts from this blog

Machine Learning Cheat Sheet (for scikit-learn)

A Wordcloud in Python

MNIST for ever....

Python things you never need: Empty lambda functions