skip to main content

Title: A Machine Learning Explainability Tutorial for Atmospheric Sciences

With increasing interest in explaining machine learning (ML) models, this paper synthesizes many topics related to ML explainability. We distinguish explainability from interpretability, local from global explainability, and feature importance versus feature relevance. We demonstrate and visualize different explanation methods, how to interpret them, and provide a complete Python package (scikit-explain) to allow future researchers and model developers to explore these explainability methods. The explainability methods include Shapley additive explanations (SHAP), Shapley additive global explanation (SAGE), and accumulated local effects (ALE). Our focus is primarily on Shapley-based techniques, which serve as a unifying framework for various existing methods to enhance model explainability. For example, SHAP unifies methods like local interpretable model-agnostic explanations (LIME) and tree interpreter for local explainability, while SAGE unifies the different variations of permutation importance for global explainability. We provide a short tutorial for explaining ML models using three disparate datasets: a convection-allowing model dataset for severe weather prediction, a nowcasting dataset for subfreezing road surface prediction, and satellite-based data for lightning prediction. In addition, we showcase the adverse effects that correlated features can have on the explainability of a model. Finally, we demonstrate the notion of evaluating model impacts of feature groups instead of individual features. Evaluating the feature groups mitigates the impacts of feature correlations and can provide a more holistic understanding of the model. All code, models, and data used in this study are freely available to accelerate the adoption of machine learning explainability in the atmospheric and other environmental sciences.

more » « less
Author(s) / Creator(s):
 ;  ;  ;  
Publisher / Repository:
American Meteorological Society
Date Published:
Journal Name:
Artificial Intelligence for the Earth Systems
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. The ability to determine whether a robot's grasp has a high chance of failing, before it actually does, can save significant time and avoid failures by planning for re-grasping or changing the strategy for that special case. Machine Learning (ML) offers one way to learn to predict grasp failure from historic data consisting of a robot's attempted grasps alongside labels of the success or failure. Unfortunately, most powerful ML models are black-box models that do not explain the reasons behind their predictions. In this paper, we investigate how ML can be used to predict robot grasp failure and study the tradeoff between accuracy and interpretability by comparing interpretable (white box) ML models that are inherently explainable with more accurate black box ML models that are inherently opaque. Our results show that one does not necessarily have to compromise accuracy for interpretability if we use an explanation generation method, such as Shapley Additive explanations (SHAP), to add explainability to the accurate predictions made by black box models. An explanation of a predicted fault can lead to an efficient choice of corrective action in the robot's design that can be taken to avoid future failures. 
    more » « less
  2. Pham, Tien ; Solomon, Latasha ; Hohil, Myron E. (Ed.)
    Explainable Artificial Intelligence (XAI) is the capability of explaining the reasoning behind the choices made by the machine learning (ML) algorithm which can help understand and maintain the transparency of the decision-making capability of the ML algorithm. Humans make thousands of decisions every day in their lives. Every decision an individual makes, they can explain the reasons behind why they made the choices that they made. Nonetheless, it is not the same in the case of ML and AI systems. Furthermore, XAI was not wideley researched until suddenly the topic was brought forward and has been one of the most relevant topics in AI for trustworthy and transparent outcomes. XAI tries to provide maximum transparency to a ML algorithm by answering questions about how models effectively came up with the output. ML models with XAI will have the ability to explain the rationale behind the results, understand the weaknesses and strengths the learning models, and be able to see how the models will behave in the future. In this paper, we investigate XAI for algorithmic trustworthiness and transparency. We evaluate XAI using some example use cases and by using SHAP (SHapley Additive exPlanations) library and visualizing the effect of features individually and cumulatively in the prediction process. 
    more » « less
  3. Abstract

    Characterization of material structure with X-ray or neutron scattering using e.g. Pair Distribution Function (PDF) analysis most often rely on refining a structure model against an experimental dataset. However, identifying a suitable model is often a bottleneck. Recently, automated approaches have made it possible to test thousands of models for each dataset, but these methods are computationally expensive and analysing the output, i.e. extracting structural information from the resulting fits in a meaningful way, is challenging. OurMachineLearning basedMotifExtractor (ML-MotEx) trains an ML algorithm on thousands of fits, and uses SHAP (SHapley Additive exPlanation) values to identify which model features are important for the fit quality. We use the method for 4 different chemical systems, including disordered nanomaterials and clusters. ML-MotEx opens for a type of modelling where each feature in a model is assigned an importance value for the fit quality based on explainable ML.

    more » « less
  4. null (Ed.)
    Feature attributions and counterfactual explanations are popular approaches to explain a ML model. The former assigns an importance score to each input feature, while the latter provides input examples with minimal changes to alter the model's predictions. To unify these approaches, we provide an interpretation based on the actual causality framework and present two key results in terms of their use. First, we present a method to generate feature attribution explanations from a set of counterfactual examples. These feature attributions convey how important a feature is to changing the classification outcome of a model, especially on whether a subset of features is necessary and/or sufficient for that change, which attribution-based methods are unable to provide. Second, we show how counterfactual examples can be used to evaluate the goodness of an attribution-based explanation in terms of its necessity and sufficiency. As a result, we highlight the complimentary of these two approaches. Our evaluation on three benchmark datasets --- Adult-Income, LendingClub, and German-Credit --- confirms the complimentary. Feature attribution methods like LIME and SHAP and counterfactual explanation methods like Wachter et al. and DiCE often do not agree on feature importance rankings. In addition, by restricting the features that can be modified for generating counterfactual examples, we find that the top-k features from LIME or SHAP are often neither necessary nor sufficient explanations of a model's prediction. Finally, we present a case study of different explanation methods on a real-world hospital triage problem. 
    more » « less
  5. Abstract. The annual area burned due to wildfires in the western United States (WUS) increased bymore than 300 % between 1984 and 2020. However, accounting for the nonlinear, spatially heterogeneous interactions between climate, vegetation, and human predictors driving the trends in fire frequency and sizes at different spatial scales remains a challenging problem for statistical fire models. Here we introduce a novel stochastic machine learning (SML) framework, SMLFire1.0, to model observed fire frequencies and sizes in 12 km × 12 km grid cells across the WUS. This framework is implemented using mixture density networks trained on a wide suite of input predictors. The modeled WUS fire frequency matches observations at both monthly (r=0.94) and annual (r=0.85) timescales, as do the monthly (r=0.90) and annual (r=0.88) area burned. Moreover, the modeled annual time series of both fire variables exhibit strong correlations (r≥0.6) with observations in 16 out of 18 ecoregions. Our ML model captures the interannual variability and the distinct multidecade increases in annual area burned for both forested and non-forested ecoregions. Evaluating predictor importance with Shapley additive explanations, we find that fire-month vapor pressure deficit (VPD) is the dominant driver of fire frequencies and sizes across the WUS, followed by 1000 h dead fuel moisture (FM1000), total monthly precipitation (Prec), mean daily maximum temperature (Tmax), and fraction of grassland cover in a grid cell. Our findings serve as a promising use case of ML techniques for wildfire prediction in particular and extreme event modeling more broadly. They also highlight the power of ML-driven parameterizations for potential implementation in fire modules of dynamic global vegetation models (DGVMs) and earth system models (ESMs). 
    more » « less