Discovery of Hierarchy in Embedding Space
Abstract: Existing learning models partition the generated representations using linear hyperplanes which form well defined groups of similar embeddings that can be uniquely mapped to a particular class. However, in practical applications, the embedding space do not form distinct boundaries to segregate the clusters. Moreover, the structure of the latent space remains obscure. As learned representations are frequently reused to reduce the inference time, it is important to analyse how semantically related classes interact among themselves in the latent space. We have proposed a cluster growing algorithm that minimises the inclusion of other classes in the embedding space to form clusters of similar representations. These clusters are overlapping to denote ambiguous embeddings that cannot be mapped to a particular class with high confidence. Later, we construct relation trees to evaluate our method with the WordNet hierarchy using phylogenetic tree comparison methods.
Submission Length: Long submission (more than 12 pages of main content)
Assigned Action Editor: ~Frederic_Sala1
Submission Number: 1644
Loading