AI Paper Atlas

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

ICE-BeeM: Identifiable Conditional Energy-Based Deep Models Based on Nonlinear ICA

Ilyes Khemakhem, Ricardo Monti, Diederik Kingma, Aapo Hyvarinen

NeurIPS 2020 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Tractable probabilistic models · 24 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 ICE-BeeM pretrained features and fit conditioning coefficients for unseen classes instead of training an unconditional EBM on those classesConvMLP-50 on MNIST · MNIST unseen classes 8-9CDSM objective: 3.64 ± 0.10 -> 2.95 ± 0.02Improved
Use ICE-BeeM pretrained features and fit conditioning coefficients for unseen classes instead of training an unconditional EBM on those classesConvMLP-90 on CIFAR10 · CIFAR10 unseen classes 8-9CDSM objective: 8.16 ± 0.05 -> 8.03 ± 0.01Improved
Train an identifiable conditional EBM rather than an unconditional EBMConvMLP-50 on FMNIST · FMNIST unseen classes 8-9classification accuracy: 56.33 ± 3.18 -> 77.07 ± 1.39Improved
Train an identifiable conditional EBM rather than an unconditional EBMConvMLP-50p on CIFAR10 · CIFAR10 unseen classes 8-9classification accuracy: 51.88 ± 1.33 -> 64.42 ± 1.09Improved
Add positive output features (ablation)ICE-BeeM ConvMLP-50 on MNIST · MNIST unseen classes 8-9classification accuracy: 76.98 ± 1.61 -> 88.46 ± 1.14Improved
Add positive output features (ablation)ICE-BeeM ConvMLP-50 on FMNIST · FMNIST unseen classes 8-9classification accuracy: 77.07 ± 1.39 -> 71.67 ± 1.85Hurt
Add positive output features (ablation)ICE-BeeM ConvMLP-50 on CIFAR10 · CIFAR10 unseen classes 8-9classification accuracy: 69.36 ± 2.23 -> 64.42 ± 1.09Hurt
Use augmented features (ablation)ICE-BeeM Unet on MNIST · MNIST unseen classes 8-9classification accuracy: 97.79 ± 0.34 -> 97.18 ± 0.5No clear effect
Increase feature dimension from 50 to 200 (ablation)ICE-BeeM ConvMLP on CIFAR10 · CIFAR10 unseen classes 8-9classification accuracy: 69.36 ± 2.23 -> 64.73 ± 1.36Hurt

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