Predictive uncertainties in classification tasks are often a consequence of
model inadequacy or insufficient training data. In popular applications, such
as image processing, we are often required to scrutinise these uncertainties by
meaningfully attributing them to input features. This helps to improve
interpretability assessments. However, there exist few effective frameworks for
this purpose. Vanilla forms of popular methods for the provision of saliency
masks, such as SHAP or integrated gradients, adapt poorly to target measures of
uncertainty. Thus, state-of-the-art tools instead proceed by creating
counterfactual or adversarial feature vectors, and assign attributions by
direct comparison to original images. In this paper, we present a novel
framework that combines path integrals, counterfactual explanations and
generative models, in order to procure attributions that contain few observable
artefacts or noise. We evidence that this outperforms existing alternatives
through quantitative evaluations with popular benchmarking methods and data
sets of varying complexity.

本文提出一种新的框架，将路径积分、反事实解释和生成模型相结合，以获得包含少量可观察到的人工构件或噪声的归因，并通过对各种复杂度的数据集的流行基准测试进行定量评估，证明这一方法优于现有的替代方法。

分类模型中预测不确定性的归因

Attribution of Predictive Uncertainties in Classification Models

In model-based reinforcement learning, planning with an imperfect model of
the environment has the potential to harm learning progress. But even when a
model is imperfect, it may still contain information that is useful for
planning. In this paper, we investigate the idea of using an imperfect model
selectively. The agent should plan in parts of the state space where the model
would be helpful but refrain from using the model where it would be harmful. An
effective selective planning mechanism requires estimating predictive
uncertainty, which arises out of aleatoric uncertainty, parameter uncertainty,
and model inadequacy, among other sources. Prior work has focused on parameter
uncertainty for selective planning. In this work, we emphasize the importance
of model inadequacy. We show that heteroscedastic regression can signal
predictive uncertainty arising from model inadequacy that is complementary to
that which is detected by methods designed for parameter uncertainty,
indicating that considering both parameter uncertainty and model inadequacy may
be a more promising direction for effective selective planning than either in
isolation.

该研究探讨了在模型不完备的情况下，通过估计预测不确定性（如起源于先验模型不充分的模型不适配性等因素）的方法，选择性地使用模型，以此提高强化学习算法的学习效果。

有限模型容量下的选择性 Dyna-style 规划

Selective Dyna-style Planning Under Limited Model Capacity

Mathematical models of physical systems are subject to many uncertainties
such as measurement errors and uncertain initial and boundary conditions. After
accounting for these uncertainties, it is often revealed that discrepancies
between the model output and the observations remain; if so, the model is said
to be inadequate. In practice, the inadequate model may be the best that is
available or tractable, and so despite its inadequacy the model may be used to
make predictions of unobserved quantities. In this case, a representation of
the inadequacy is necessary, so the impact of the observed discrepancy can be
determined. We investigate this problem in the context of chemical kinetics and
propose a new technique to account for model inadequacy that is both
probabilistic and physically meaningful. A stochastic inadequacy operator
$\mathcal{S}$ is introduced which is embedded in the ODEs describing the
evolution of chemical species concentrations and which respects certain
physical constraints such as conservation laws. The parameters of $\mathcal{S}$
are governed by probability distributions, which in turn are characterized by a
set of hyperparameters. The model parameters and hyperparameters are calibrated
using high-dimensional hierarchical Bayesian inference. We apply the method to
a typical problem in chemical kinetics---the reaction mechanism of hydrogen
combustion.

本文提出一种概率的物理意义下的随机不足算子，用于化学动力学反应机理的不充分模型，通过高维层级贝叶斯推断对模型参数及超参数进行校准，以解决数学模型适应性差的问题。