Shared Kernel Information Embedding for Discriminative Inference

Latent Variable Models (LVM), like the Shared-GPLVM and the Spectral Latent Variable Model, help mitigate over- fitting when learning discriminative methods from small or moderately sized training sets. Nevertheless, existing meth- ods suffer from several problems: 1) complexity; 2) the lack of explicit mappings to and from the latent space; 3) an inability to cope with multi-modality; and 4) the lack of a well-defined density over the latent space. We pro- pose a LVM called the Shared Kernel Information Em- bedding (sKIE). It defines a coherent density over a latent space and multiple input/output spaces (e.g., image features and poses), and it is easy to condition on a latent state, or on combinations of the input/output states. Learning is quadratic, and it works well on small datasets. With datasets too large to learn a coherent global model, one can use sKIE to learn local online models. sKIE permits missing data during inference, and partially labelled data durin...

• Type: Conference
• Year: 2009
• Where: CVPR
• Authors: David J. Fleet, Leonid Sigal, Roland Memisevic
• Keywords: Computer Vision, CVPR 2009, Discriminative Inference