We unified some latent block models by proposing a flexible ELBM that is extended to SELBM to address the sparse problem by revealing a diagonal structure from sparse datasets. This leads to obtain ...