Manifold alignment preserving global geometry
Chang Wang, Sridhar Mahadevan · 2013
This paper proposes a novel algorithm for man-ifold alignment preserving global geometry. This approach constructs mapping functions that project data instances from different input domains to a new lower-dimensional space, simultaneously matching the instances in correspondence and pre-serving global distances between instances within the original domains. In contrast to previous ap-proaches, which are largely based on preserving lo-cal geometry, the proposed approach is suited to applications where the global manifold geometry needs to be respected. We evaluate the effective-ness of our algorithm for transfer learning in two real-world cross-lingual information retrieval tasks. 1