Research map / Architectures and efficiency / RNNs and normalization
FiLM-Ensemble: Probabilistic Deep Learning via Feature-wise Linear Modulation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use FiLM modulation parameters to form ensemble members while sharing backbone weights | Deep neural network ensembles · CIFAR-100, ResNet-18 | ECE: 0.041 $\pm 0.002$ -> 0.038 $\pm 0.000$ | Helped |
| Use FiLM modulation parameters to form ensemble members while sharing backbone weights | Deep neural network ensembles · OOD retinal Glaucoma / REFUGE | AUROC: $78.22 \pm 0.1$ -> $79.85 \pm 0.2$ | Helped |
| Increase ensemble size | FiLM-Ensemble · CIFAR-10 | ECE | Helped |
| Increase ensemble size | BatchEnsemble · CIFAR-10 | accuracy | Hurt |
| Increase initialization gain ρ (ablation) | FiLM-Ensemble · 6mA-rice-Lv | ECE | Helped |
| Increase initialization gain ρ (ablation) | FiLM-Ensemble · 6mA-rice-Lv | accuracy: <0.25 percent points | Hurt |
| Increase ensemble size | FiLM-Ensemble · Retinal Glaucoma | accuracy: 86.3 -> 87.8 | Helped |
| Increase ensemble size | Deep Ensemble · Retinal Glaucoma | ECE: $0.041 \pm 0.002$ -> $0.078 \pm 0.002$ -> $0.091 \pm 0.004$ -> $0.066 \pm 0.003$ | Mixed |
| Increase ensemble size | FiLM-Ensemble · OOD retinal Glaucoma / REFUGE | AUROC: $77.02 \pm 0.1$ -> $77.92 \pm 0.2 $ -> $79.43 \pm 0.1$ -> $79.85 \pm 0.2$ | Helped |
Closest papers by meaning
- Neural Ensemble Search for Uncertainty Estimation and Dataset Shift (NeurIPS 2021 · similarity 0.88)
- Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles (NeurIPS 2017 · similarity 0.88)
- Packed Ensembles for efficient uncertainty estimation (ICLR 2023 · similarity 0.88)
- Hyperparameter Ensembles for Robustness and Uncertainty Quantification (NeurIPS 2020 · similarity 0.88)
- Uncertainty Quantification and Deep Ensembles (NeurIPS 2021 · similarity 0.87)
- Bayesian Exploration of Pre-trained Models for Low-shot Image Classification (CVPR 2024 · similarity 0.85)
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.