1 Benchmark算法
算法 | Paper |
---|---|
Bayes by Backprop | “Weight uncertainty in neural networks.” ICML, 2015.[Link] |
Monte Carlo Dropout | “Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning.” ICML, 2016.[Link] |
Deep Ensembles | “Simple and scalable predictive uncertainty estimation using deep ensembles.” NeurIPS, 2017.[Link] |
Calibrated Regression | “Accurate Uncertainties for Deep Learning Using Calibrated Regression.” ICML, 2018.[Link] |
Probabilistic Backpropagation | “Probabilistic backpropagation for scalable learning of bayesian neural networks.” ICML, 2015.[Link] |
Ensemble MC Dropout | "Understanding Measures of Uncertainty for Adversarial Example Detection."UAI, 2018.[Link] |
Prior Networks | “Predictive uncertainty estimation via prior networks.” NeurIPS, 2018.[Link] |
Deterministic Uncertainty Quantification | “Uncertainty Estimation Using a Single Deep Deterministic Neural Network.” ICML, 2020.[Link] |
Simultaneous Quantile Regression | “Single-Model Uncertainties for Deep Learning.” NeurIPS, 2019.[Link] |
Approximated Variance Propagation | “Sampling-free Epistemic Uncertainty Estimation Using Approximated Variance Propagation.” ICCV, 2019.[Link] |
SDE-Net | “SDE-Net: Equipping Deep Neural Networks with Uncertainty Estimates.” ICML, 2020. [Link] |
Deterministic Variational Inference | “Deterministic Variational Inference for Robust Bayesian Neural Networks” ICLR, 2019. [Link] |
2 任务和数据集
-
简单回归问题探索不确定性的意义
- Boston housing、Concrete、Energy、Kin8nm、Naval propulsion plant、Power plant、Protein、Wine、Yacht、Year Prediction MSD(主要来源于Kaggle,涉及真实世界中真实数据类似一段时间的股票,主要用于回归拟合)
-
量化各种任务中的不确定性
- 热布局数据集
- Cityscapes、CamVid、NYUv2(语义分割)、
-
医疗图像(包括分类,分割;能否可靠的指导医学实践?)
-
自动驾驶(自动驾驶系统识别出的前方物体是否可信?)
-
主动学习(什么样的样本需要被标注?)
-
强化学习(如何权衡探索与利用?)
-
异常样本OOD检测(如何检测?)
- MNIST、CIFAR10、CIFAR100、SVHN、TinyImageNet、LSUN、Fashion MNIST、KMNIST、 SEMEION、STL-10
-
对抗样本检测(如何检测?)
-
理解模型、数据(我们知道什么又不知道什么?)
-
模型鲁棒研究
-
…