The Responsibility Problem in Neural Networks with Unordered Targets
This addresses a foundational issue in neural network design for unordered data, but it is incremental as it builds on prior work by extending the proof from a single discontinuity to uncountably many.
The paper tackled the responsibility problem in neural networks with unordered targets by proving that discontinuities are uncountably infinite, motivating further research in this area.
We discuss the discontinuities that arise when mapping unordered objects to neural network outputs of fixed permutation, referred to as the responsibility problem. Prior work has proved the existence of the issue by identifying a single discontinuity. Here, we show that discontinuities under such models are uncountably infinite, motivating further research into neural networks for unordered data.