Learning to Bound: Monotonic, Calibrated Neural Upper Bounds for MPE Inference
Learns monotonic calibrated upper bounds to speed branch-and-bound for probabilistic inference.
Explainer · on the wire
The explainer for this story is still on the wire.
The claim, the mechanism, and the figures land here once generated.
Source · arXiv