Computational efficiency for active learning

[dongwookim-ml]

Problem:

• Current sequential Monte-Carlo is still inapplicable to large scale knowledge graphs.

Possible approaches:

• Random sampling on subset of entities and relations.
• Hierarchical sampling on entities and relations.

[dongwookim-ml]

• Apply effective sample size for each triple not for each particle.