AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Evidential Softmax for Sparse Multimodal Distributions in Deep Generative Models

Phil Chen, Masha Itkina, Ransalu Senanayake, Mykel Kochenderfer

NeurIPS 2021 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Tractable probabilistic models · 0 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
Train with sparse, multimodal Ev-softmax rather than Entmax-1.5Semi-supervised VAE · MNISTaccuracy (%): 97.20 ± .001 -> 97.30 ± .001Improved
Use Ev-softmax rather than SparsemaxSemi-supervised VAE · MNISTdecoder calls: 1.01 ± 0.01 -> 1.64 ± 0.01Hurt
Train the PixelCNN prior with Ev-softmax rather than SparsemaxVQ-VAE · tinyImageNettop-5 accuracy: 40.0 ± 0.2 -> 43.6 ± 0.3Improved
Train the PixelCNN prior with Ev-softmax rather than SparsemaxVQ-VAE · tinyImageNettop-10 accuracy: 52.0 ± 0.2 -> 55.6 ± 0.2Improved
Use Ev-softmax rather than SparsemaxVQ-VAE · tinyImageNetaverage number of nonzero latent classes: 46 ± 0.3 -> 77 ± 0.2Mixed
Use Ev-softmax rather than SoftmaxTransformer self-attention · IWSLT 2014 EN→DEBLEU: 29.2 ± 0.06 -> 29.4 ± 0.05Improved
Use sparse Ev-softmax attention rather than attention over the full sourceTransformer self-attention · IWSLT 2014 EN→DEattended source words: 39.5 ± 11.5 -> 8.2 ± 1.3Mixed
Train the CVAE with Ev-softmax rather than SoftmaxCVAE · MNIST even/odd generationWasserstein distanceImproved
Train the CVAE with Ev-softmax rather than Sparsemax or Entmax-1.5CVAE · MNIST even/odd generationWasserstein distanceImproved
Train with sparse and multimodal outputs rather than applying evidential sparsification only after trainingSemi-supervised VAE · MNISTaccuracy (%): 96.90 ± .001 -> 97.30 ± .001Improved
Train the prior with sparse normalization rather than SoftmaxVQ-VAE PixelCNN prior · tinyImageNettop-5 accuracy: 38.4 ± 0.2 -> 43.6 ± 0.3Improved
Retain more attended source words with Ev-softmax rather than SparsemaxTransformer self-attention · IWSLT 2014 EN→DEBLEU: 29.0 ± 0.05 -> 29.4 ± 0.05Improved

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