AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

DropMax: Adaptive Variational Softmax

Hae Beom Lee, Juho Lee, Saehoon Kim, Eunho Yang, Sung Ju Hwang

NeurIPS 2018 · Label-efficient and robust learning › Recognition and long-tail · approach: Learning from unlabeled and positive-unlabeled data · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use input-adaptive stochastic class dropout with variational inference instead of Base SoftmaxCNN · MNIST-1Ktest error (%): 7.09 -> 5.32Improved
Use input-adaptive stochastic class dropout with variational inference instead of Base SoftmaxCNN · MNIST-5Ktest error (%): 2.13 -> 1.64Improved
Use input-adaptive stochastic class dropout with variational inference instead of Base SoftmaxCNN · MNIST-55Ktest error (%): 0.65 -> 0.59Improved
Use input-adaptive stochastic class dropout with variational inference instead of Base SoftmaxResNet-34 · CIFAR-10test error (%): 7.90 -> 7.67Improved
Use input-adaptive stochastic class dropout with variational inference instead of Base SoftmaxResNet-34 · CIFAR-100test error (%): 30.60 -> 29.87Improved
Use input-adaptive stochastic class dropout with variational inference instead of Base SoftmaxResNet-18 · AWAtest error (%): 30.29 -> 26.91Improved
Use input-adaptive stochastic class dropout with variational inference instead of Base SoftmaxResNet-18 · CUB-200-2011test error (%): 48.84 -> 41.07Improved
Replace label-informed approximate posterior with q=p (ablation)DropMax · CUB-200-2011test error (%): 41.07 -> 42.08Hurt
Replace stochastic adaptive class dropout with supervised deterministic attention (ablation)DropMax · CUB-200-2011test error (%): 41.07 -> 47.35Hurt
Use DropMax instead of Base Softmax on fine-grained datasetsResNet-18 · AWA and CUB-200-2011test error (%): 30.29 -> 26.91; 48.84 -> 41.07Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy