Analternativeapproach isto predict instead adistributionp(ˆy|x) = Φˆy(x) over possible values of the annotationy. Theannotators are shown a set of points sampled randomly and uniformly over one of predefined body parts of aperson inan image.

Optimization is an integral part of most machine learning systems and most numerical optimization schemes relyonthecomputation ofderivatives.

Many works have studied neural execution in different domains before [45, 23, 24, 31, 33, 42]. With the rapid development of GNNs in graph representation learning, learning graph algorithms with GNNs has attracted researchers' attention [39, 38, 41]. These works exploit GNNs to approximate certain classes of graph algorithms, such as parallel algorithms (e.g., Breadth-First-Search) and sequential algorithms (e.g., Dijkstra).

However,computing extendedpersistent homologysummaries remainsslowfor large and dense graphs and can be aserious bottleneck for the learning pipeline.

However, these decisions may be imperfect due to limited experience, implicit biases, or faulty probabilistic reasoning.

Neural Information Processing Systems http://nips.cc/

Neural Information Processing Systems http://nips.cc/

Wepropose alearning algorithm which accounts for potential biases held by external decision-makers in a system.