Unimodal probability distributions for deep ordinal classification
This addresses a specific problem in deep ordinal classification for researchers and practitioners, but it is incremental as it builds on existing methods with a straightforward constraint.
The paper tackled the issue of undesirable properties in probability distributions for ordinal classification by proposing a technique to enforce unimodality using Poisson and binomial distributions, achieving promising results on two large image datasets.
Probability distributions produced by the cross-entropy loss for ordinal classification problems can possess undesired properties. We propose a straightforward technique to constrain discrete ordinal probability distributions to be unimodal via the use of the Poisson and binomial probability distributions. We evaluate this approach in the context of deep learning on two large ordinal image datasets, obtaining promising results.