ICML 2011 workshop on unsupervised and transfer learning

Use of Representations in High Dimensional Spaces for Unsupervised and Transfer Learning Challenge

Mehreen Saeed (aliphlaila team, Fourth place phase 2 UTL challenge)
FAST, National University of Computer and Emerging Sciences
Lahore, Pakistan.

This paper describes our work for the unsupervised and transfer learning challenge. The goal of the challenge was to nd representations, of data, which are linearly separable for a variety of multiclass problems. The datasets chosen for the challenge belong to different domains. We show how manifold learning and simple similarity kernels can be used to get good results on these datasets. Our entry had a global ranking of 4 in the second phase of this challenge and results on some datasets are comparable to the fist ranked entries.