Unimodal regularisation based on beta distribution for deep ordinal regression

作者：

Highlights：

• Unimodal regularisation for cross-entropy loss using the beta distribution.

• Comparison with other regularisation methods based on other distributions.

• Ordinal output scheme based on the stick-breaking process.

• Application to convolutional neural networks and several benchmark datasets.

摘要

•Unimodal regularisation for cross-entropy loss using the beta distribution.•Comparison with other regularisation methods based on other distributions.•Ordinal output scheme based on the stick-breaking process.•Application to convolutional neural networks and several benchmark datasets.

论文关键词：Ordinal regression,Unimodal distribution,Convolutional network,Beta distribution,Stick-breaking

论文评审过程：Received 3 March 2021, Revised 6 September 2021, Accepted 7 September 2021, Available online 10 September 2021, Version of Record 20 September 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108310