Secure Semi-supervised Vector Quantization for Dissimilarity Data

The amount and complexity of data increase rapidly, however, due to time and cost constrains, only few of them are fully labeled. In this context non-vectorial relational data given by pairwise (dis)similarities without explicit vectorial representation, like score-values in sequences alignments, are particularly challenging. Existing semi-supervised learning (SSL) algorithms focus on vectorial data given in Euclidean space. In this paper we extend a prototype-based classifier for dissimilarity data to non i.i.d. semi-supervised tasks. Using conformal prediction the ’secure region’ of unlabeled data can be used to improve the trained model based on labeled data while adapting the model complexity to cover the ’insecure region’ of labeled data. The proposed method is evaluated on some benchmarks from the SSL domain.