Research map / Adversarial robustness and security / Certified and empirical robustness
Lightweight Probabilistic Deep Networks
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 ADF uncertainty propagation for optical-flow regression | FlowNetS · Sintel clean test | EPE: 7.66 -> 7.46 | Helped |
| Use ADF uncertainty propagation for optical-flow regression | FlowNetS · Sintel final test | EPE: 8.53 -> 8.53 | No clear effect |
| Use probabilistic output layers for optical-flow regression | FlowNetS · Sintel final test | EPE: 8.53 -> 8.30 | Helped |
| Use probabilistic output layers for optical-flow regression | FlowNetS · FlyingChairs test | EPE: 2.38 -> 2.15 | Helped |
| Use a Dirichlet output with conditional likelihood training | All-CNN-C · CIFAR10 test | accuracy (%): 90.62 -> 91.87 | Helped |
| Use a Dirichlet output with conditional likelihood training | LeNet · MNIST test | accuracy (%): 99.39 -> 99.48 | Helped |
| Use ADF with a Dirichlet output and conditional likelihood | All-CNN-C · CIFAR10 test | accuracy (%): 90.62 -> 91.51 | Helped |
| Use a Dirichlet output with conditional likelihood training | All-CNN-C · CIFAR10 FGSM, ǫ=0.01 | accuracy (%): 46.66 -> 66.49 | Helped |
| Use a Dirichlet output with conditional likelihood training | All-CNN-C · CIFAR10 FGSM, ǫ=0.05 | accuracy (%): 35.79 -> 46.26 | Helped |
| Use a Dirichlet output with conditional likelihood training | All-CNN-C · CIFAR10 FGSM, ǫ=0.1 | accuracy (%): 23.99 -> 23.20 | Helped |
| Replace the softmax with a Dirichlet output | CIFAR10 classification · CIFAR10 test | uncertainty calibration | Helped |
| Use test-time dropout sampling for optical-flow regression | FlowNetS · Sintel clean test | EPE: 7.66 -> 7.65 | No clear effect |
Closest papers by meaning
- Depth Uncertainty in Neural Networks (NeurIPS 2020 · similarity 0.89)
- Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles (NeurIPS 2017 · similarity 0.88)
- Bayesian Uncertainty Estimation for Batch Normalized Deep Networks (ICML 2018 · similarity 0.87)
- Deep Probabilistic Supervision for Image Classification (NeurIPS 2026 · similarity 0.87)
- Make Me a BNN: A Simple Strategy for Estimating Bayesian Uncertainty from Pre-trained Models (CVPR 2024 · similarity 0.86)
- SURE: SUrvey REcipes for building reliable and robust deep networks (CVPR 2024 · similarity 0.86)
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.