Selection of most representative training examples and incremental generation of VL₁ hypotheses | lit.salon