Supervised Manifold Learning with Incremental Stochastic Embeddings
Oliver Krämer · The European Symposium on Artificial Neural Networks · 2015
In this paper, we introduce an incremental dimensionality reduction approach for labeled data. The algorithm incrementally sam- ples in latent space and chooses a solution that minimizes the nearest neighbor classication error taking into account label information. We introduce and compare two optimization approaches to generate super- vised embeddings, i.e., an incremental solution construction method and a re-embedding approach. Both methods have in common that the objec- tive is to minimize the nearest neighbor classication error computed in the low-dimensional space. The resulting embedding is a surrogate of the high-dimensional labeled set. The set allows conclusions about the data set structure and can be used as preprocessing step for classication of labeled patterns.