The black hole imaging problem is non-convex and highly ill-posed due to severe noise corruption and measurement sparsity.

This makes CDE particularly useful in critical application domains. However, interpretable CDE methods are understudied.

We establish a new theoretical framework for learning under multi-class, instance-dependent label noise. This framework casts learning with label noise as a form of domain adaptation, in particular, domain adaptation under posterior drift.

Utilizing mutual learning can effectively enhance the performance of peer BNNs.

Critical in this context is the prior knowledge accumulated from related tasks.

Machine learning (ML) models have been widely used to make predictions.