AI Paper Atlas

Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes

VICE: Variational Interpretable Concept Embeddings

Lukas Muttenthaler, Charles Yang Zheng, Patrick McClure, Robert A. Vandermeulen, Martin N Hebart, Francisco Pereira

NeurIPS 2022 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Bayesian view of meta-learning

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 VICE variational spike-and-slab embeddings instead of SPoSEobject embedding model · THINGS full training datatriplet accuracy: 0.637 -> 0.638No clear effect
use VICE instead of SPoSEobject embedding model · THINGS full training dataKL divergence: 0.103 -> 0.100No clear effect
use VICE instead of SPoSEobject embedding model · Adjectives testtriplet accuracy: 0.562 -> 0.559No clear effect
use VICE instead of SPoSEobject embedding model · Food testtriplet accuracy: 0.698 -> 0.693No clear effect
train on only 5% and 10% of THINGS dataVICE vs SPoSE · THINGS small subsetsaccuracy and KLImproved
use VICE instead of SPoSE on full THINGSembedding model · THINGSnumber of selected dimensions: 52 -> 44Improved
use VICE instead of SPoSEembedding model · Foodnumber of selected dimensions: 16 -> 5Improved
use VICE instead of SPoSEembedding model · Adjectivesnumber of selected dimensions: 22 -> 21No clear effect
use VICE instead of SPoSEembedding model · THINGSdimension reproducibility: 81.30 -> 87.01%Improved
use VICE instead of SPoSEembedding model · Fooddimension reproducibility: 62.88 -> 87.38%Improved
use VICE instead of SPoSEembedding model · Adjectivesdimension reproducibility: 71.64 -> 76.76%Improved
use default hyperparameters instead of grid-searched best (ablation)VICE · all three datasetsaccuracy: within 0.015No clear effect

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