In this review, we examine the problem of designing interpretable and explainable machine learning models. Interpretability and explainability lie at the core of many machine learning and statistical applications in medicine, economics, law, and natural sciences. Although interpretability and explainability have escaped a clear universal definition, many techniques motivated by these properties have been developed over the recent 30 years with the focus currently shifting towards deep learning methods. In this review, we emphasise the divide between interpretability and explainability and illustrate these two different research directions with concrete examples of the state-of-the-art. The review is intended for a general machine learning audience with interest in exploring the problems of interpretation and explanation beyond logistic regression or random forest variable importance. This work is not an exhaustive literature survey, but rather a primer focusing selectively on certain lines of research which the authors found interesting or informative.
Prototype-based methods use interpretable representations to address the black-box nature of deep learning models, in contrast to post-hoc explanation methods that only approximate such models. We propose the Neural Prototype Tree (ProtoTree), an intrinsically interpretable deep learning method for fine-grained image recognition. ProtoTree combines prototype learning with decision trees, and thus results in a globally interpretable model by design. Additionally, ProtoTree can locally explain a single prediction by outlining a decision path through the tree. Each node in our binary tree contains a trainable prototypical part. The presence or absence of this learned prototype in an image determines the routing through a node. Decision making is therefore similar to human reasoning: Does the bird have a red throat? And an elongated beak? Then it's a hummingbird! We tune the accuracy-interpretability trade-off using ensemble methods, pruning and binarizing. We apply pruning without sacrificing accuracy, resulting in a small tree with only 8 learned prototypes along a path to classify a bird from 200 species. An ensemble of 5 ProtoTrees achieves competitive accuracy on the CUB-200- 2011 and Stanford Cars data sets. Code is available at https://github.com/M-Nauta/ProtoTree
Self-organized pattern behavior is ubiquitous throughout nature, from fish schooling to collective cell dynamics during organism development. Qualitatively these patterns display impressive consistency, yet variability inevitably exists within pattern-forming systems on both microscopic and macroscopic scales. Quantifying variability and measuring pattern features can inform the underlying agent interactions and allow for predictive analyses. Nevertheless, current methods for analyzing patterns that arise from collective behavior only capture macroscopic features, or rely on either manual inspection or smoothing algorithms that lose the underlying agent-based nature of the data. Here we introduce methods based on topological data analysis and interpretable machine learning for quantifying both agent-level features and global pattern attributes on a large scale. Because the zebrafish is a model organism for skin pattern formation, we focus specifically on analyzing its skin patterns as a means of illustrating our approach. Using a recent agent-based model, we simulate thousands of wild-type and mutant zebrafish patterns and apply our methodology to better understand pattern variability in zebrafish. Our methodology is able to quantify the differential impact of stochasticity in cell interactions on wild-type and mutant patterns, and we use our methods to predict stripe and spot statistics as a function of varying cellular communication. Our work provides a new approach to automatically quantifying biological patterns and analyzing agent-based dynamics so that we can now answer critical questions in pattern formation at a much larger scale.
The use of convolutional neural networks (CNNs) for classification tasks has become dominant in various medical imaging applications. At the same time, recent advances in interpretable machine learning techniques have shown great potential in explaining classifiers' decisions. Layer-wise relevance propagation (LRP) has been introduced as one of these novel methods that aim to provide visual interpretation for the network's decisions. In this work we propose the application of 3D CNNs with LRP for the first time for neonatal T2-weighted magnetic resonance imaging (MRI) data analysis. Through LRP, the decisions of our trained classifier are transformed into heatmaps indicating each voxel's relevance for the outcome of the decision. Our resulting LRP heatmaps reveal anatomically plausible features in distinguishing preterm neonates from term ones.
The human electroencephalogram (EEG) of sleep undergoes profound changes with age. These changes can be conceptualized as "brain age", which can be compared to an age norm to reflect the deviation from normal aging process. Here, we develop an interpretable machine learning model to predict brain age based on two large sleep EEG datasets: the Massachusetts General Hospital sleep lab dataset (MGH, N = 2,621) covering age 18 to 80; and the Sleep Hearth Health Study (SHHS, N = 3,520) covering age 40 to 80. The model obtains a mean absolute deviation of 8.1 years between brain age and chronological age in the healthy participants in the MGH dataset. As validation, we analyze a subset of SHHS containing longitudinal EEGs 5 years apart, which shows a 5.5 years difference in brain age. Participants with neurological and psychiatric diseases, as well as diabetes and hypertension medications show an older brain age compared to chronological age. The findings raise the prospect of using sleep EEG as a biomarker for healthy brain aging.
COVID-19 is not a universal killer. We study the spread of COVID-19 at the county level for the United States up until the 15$^{th}$ of August, 2020. We show that the prevalence of the disease and the death rate are correlated with the local socio-economic conditions often going beyond local population density distributions, especially in rural areas. We correlate the COVID-19 prevalence and death rate with data from the US Census Bureau and point out how the spreading patterns of the disease show asymmetries in urban and rural areas separately and are preferentially affecting the counties where a large fraction of the population is non-white. Our findings can be used for more targeted policy building and deployment of resources for future occurrence of a pandemic due to SARS-CoV-2. Our methodology, based on interpretable machine learning and game theory, can be extended to study the spread of other diseases.
Explainable artificial intelligence (XAI) methods are currently evaluated with approaches mostly originated in interpretable machine learning (IML) research that focus on understanding models such as comparison against existing attribution approaches, sensitivity analyses, gold set of features, axioms, or through demonstration of images. There are problems with these methods such as that they do not indicate where current XAI approaches fail to guide investigations towards consistent progress of the field. They do not measure accuracy in support of accountable decisions, and it is practically impossible to determine whether one XAI method is better than the other or what the weaknesses of existing models are, leaving researchers without guidance on which research questions will advance the field. Other fields usually utilize ground-truth data and create benchmarks. Data representing ground-truth explanations is not typically used in XAI or IML. One reason is that explanations are subjective, in the sense that an explanation that satisfies one user may not satisfy another. To overcome these problems, we propose to represent explanations with canonical equations that can be used to evaluate the accuracy of XAI methods. The contributions of this paper include a methodology to create synthetic data representing ground-truth explanations, three data sets, an evaluation of LIME using these data sets, and a preliminary analysis of the challenges and potential benefits in using these data to evaluate existing XAI approaches. Evaluation methods based on human-centric studies are outside the scope of this paper.
The definition of Linear Symmetry-Based Disentanglement (LSBD) formalizes the notion of linearly disentangled representations, but there is currently no metric to quantify LSBD. Such a metric is crucial to evaluate LSBD methods and to compare to previous understandings of disentanglement. We propose $\mathcal{D}_\mathrm{LSBD}$, a mathematically sound metric to quantify LSBD, and provide a practical implementation for $\mathrm{SO}(2)$ groups. Furthermore, from this metric we derive LSBD-VAE, a semi-supervised method to learn LSBD representations. We demonstrate the utility of our metric by showing that (1) common VAE-based disentanglement methods don't learn LSBD representations, (2) LSBD-VAE as well as other recent methods can learn LSBD representations, needing only limited supervision on transformations, and (3) various desirable properties expressed by existing disentanglement metrics are also achieved by LSBD representations.
In this paper, we explore connections between interpretable machine learning and learning theory through the lens of local approximation explanations. First, we tackle the traditional problem of performance generalization and bound the test-time accuracy of a model using a notion of how locally explainable it is. Second, we explore the novel problem of explanation generalization which is an important concern for a growing class of finite sample-based local approximation explanations. Finally, we validate our theoretical results empirically and show that they reflect what can be seen in practice.
Identifying anomalies in the fuel consumption of the vehicles of a fleet is a crucial aspect for optimizing consumption and reduce costs. However, this information alone is insufficient, since fleet operators need to know the causes behind anomalous fuel consumption. We combine unsupervised anomaly detection techniques, domain knowledge and interpretable Machine Learning models for explaining potential causes of abnormal fuel consumption in terms of feature relevance. The explanations are used for generating recommendations about fuel optimization, that are adjusted according to two different user profiles: fleet managers and fleet operators. Results are evaluated over real-world data from telematics devices connected to diesel and petrol vehicles from different types of industrial fleets. We measure the proposal regarding model performance, and using Explainable AI metrics that compare the explanations in terms of representativeness, fidelity, stability, contrastiveness and consistency with apriori beliefs. The potential fuel reductions that can be achieved is round 35%.
Decision trees (DTs) epitomize what have become to be known as interpretable machine learning (ML) models. This is informally motivated by paths in DTs being often much smaller than the total number of features. This paper shows that in some settings DTs can hardly be deemed interpretable, with paths in a DT being arbitrarily larger than a PI-explanation, i.e. a subset-minimal set of feature values that entails the prediction. As a result, the paper proposes a novel model for computing PI-explanations of DTs, which enables computing one PI-explanation in polynomial time. Moreover, it is shown that enumeration of PI-explanations can be reduced to the enumeration of minimal hitting sets. Experimental results were obtained on a wide range of publicly available datasets with well-known DT-learning tools, and confirm that in most cases DTs have paths that are proper supersets of PI-explanations.
We present a brief history of the field of interpretable machine learning (IML), give an overview of state-of-the-art interpretation methods, and discuss challenges. Research in IML has boomed in recent years. As young as the field is, it has over 200 years old roots in regression modeling and rule-based machine learning, starting in the 1960s. Recently, many new IML methods have been proposed, many of them model-agnostic, but also interpretation techniques specific to deep learning and tree-based ensembles. IML methods either directly analyze model components, study sensitivity to input perturbations, or analyze local or global surrogate approximations of the ML model. The field approaches a state of readiness and stability, with many methods not only proposed in research, but also implemented in open-source software. But many important challenges remain for IML, such as dealing with dependent features, causal interpretation, and uncertainty estimation, which need to be resolved for its successful application to scientific problems. A further challenge is a missing rigorous definition of interpretability, which is accepted by the community. To address the challenges and advance the field, we urge to recall our roots of interpretable, data-driven modeling in statistics and (rule-based) ML, but also to consider other areas such as sensitivity analysis, causal inference, and the social sciences.
A method for the local and global interpretation of a black-box model on the basis of the well-known generalized additive models is proposed. It can be viewed as an extension or a modification of the algorithm using the neural additive model. The method is based on using an ensemble of gradient boosting machines (GBMs) such that each GBM is learned on a single feature and produces a shape function of the feature. The ensemble is composed as a weighted sum of separate GBMs resulting a weighted sum of shape functions which form the generalized additive model. GBMs are built in parallel using randomized decision trees of depth 1, which provide a very simple architecture. Weights of GBMs as well as features are computed in each iteration of boosting by using the Lasso method and then updated by means of a specific smoothing procedure. In contrast to the neural additive model, the method provides weights of features in the explicit form, and it is simply trained. A lot of numerical experiments with an algorithm implementing the proposed method on synthetic and real datasets demonstrate its efficiency and properties for local and global interpretation.
Explainable AI is an emerging field providing solutions for acquiring insights into automated systems' rationale. It has been put on the AI map by suggesting ways to tackle key ethical and societal issues. Existing explanation techniques are often not comprehensible to the end user. Lack of evaluation and selection criteria also makes it difficult for the end user to choose the most suitable technique. In this study, we combine logic-based argumentation with Interpretable Machine Learning, introducing a preliminary meta-explanation methodology that identifies the truthful parts of feature importance oriented interpretations. This approach, in addition to being used as a meta-explanation technique, can be used as an evaluation or selection tool for multiple feature importance techniques. Experimentation strongly indicates that an ensemble of multiple interpretation techniques yields considerably more truthful explanations.
The black-box nature of machine learning models hinders the deployment of some high-accuracy models in medical diagnosis. It is risky to put one's life in the hands of models that medical researchers do not fully understand. However, through model interpretation, black-box models can promptly reveal significant biomarkers that medical practitioners may have overlooked due to the surge of infected patients in the COVID-19 pandemic. This research leverages a database of 92 patients with confirmed SARS-CoV-2 laboratory tests between 18th Jan. 2020 and 5th Mar. 2020, in Zhuhai, China, to identify biomarkers indicative of severity prediction. Through the interpretation of four machine learning models, decision tree, random forests, gradient boosted trees, and neural networks using permutation feature importance, Partial Dependence Plot (PDP), Individual Conditional Expectation (ICE), Accumulated Local Effects (ALE), Local Interpretable Model-agnostic Explanations (LIME), and Shapley Additive Explanation (SHAP), we identify an increase in N-Terminal pro-Brain Natriuretic Peptide (NTproBNP), C-Reaction Protein (CRP), and lactic dehydrogenase (LDH), a decrease in lymphocyte (LYM) is associated with severe infection and an increased risk of death, which is consistent with recent medical research on COVID-19 and other research using dedicated models. We further validate our methods on a large open dataset with 5644 confirmed patients from the Hospital Israelita Albert Einstein, at S\~ao Paulo, Brazil from Kaggle, and unveil leukocytes, eosinophils, and platelets as three indicative biomarkers for COVID-19.
Machine learning has emerged as a powerful approach in materials discovery. Its major challenge is selecting features that create interpretable representations of materials, useful across multiple prediction tasks. We introduce an end-to-end machine learning model that automatically generates descriptors that capture a complex representation of a material's structure and chemistry. This approach builds on computational topology techniques (namely, persistent homology) and word embeddings from natural language processing. It automatically encapsulates geometric and chemical information directly from the material system. We demonstrate our approach on multiple nanoporous metal-organic framework datasets by predicting methane and carbon dioxide adsorption across different conditions. Our results show considerable improvement in both accuracy and transferability across targets compared to models constructed from the commonly-used, manually-curated features, consistently achieving an average 25-30% decrease in root-mean-squared-deviation and an average increase of 40-50% in R2 scores. A key advantage of our approach is interpretability: Our model identifies the pores that correlate best to adsorption at different pressures, which contributes to understanding atomic-level structure--property relationships for materials design.
Health insurance companies cover half of the United States population through commercial employer-sponsored health plans and pay 1.2 trillion US dollars every year to cover medical expenses for their members. The actuary and underwriter roles at a health insurance company serve to assess which risks to take on and how to price those risks to ensure profitability of the organization. While Bayesian hierarchical models are the current standard in the industry to estimate risk, interest in machine learning as a way to improve upon these existing methods is increasing. Lumiata, a healthcare analytics company, ran a study with a large health insurance company in the United States. We evaluated the ability of machine learning models to predict the per member per month cost of employer groups in their next renewal period, especially those groups who will cost less than 95\% of what an actuarial model predicts (groups with "concession opportunities"). We developed a sequence of two models, an individual patient-level and an employer-group-level model, to predict the annual per member per month allowed amount for employer groups, based on a population of 14 million patients. Our models performed 20\% better than the insurance carrier's existing pricing model, and identified 84\% of the concession opportunities. This study demonstrates the application of a machine learning system to compute an accurate and fair price for health insurance products and analyzes how explainable machine learning models can exceed actuarial models' predictive accuracy while maintaining interpretability.
We introduce a machine learning approach to predict chronic homelessness from de-identified client shelter records drawn from a commonly used Canadian homelessness management information system. Using a 30-day time step, a dataset for 6521 individuals was generated. Our model, HIFIS-RNN-MLP, incorporates both static and dynamic features of a client's history to forecast chronic homelessness 6 months into the client's future. The training method was fine-tuned to achieve a high F1-score, giving a desired balance between high recall and precision. Mean recall and precision across 10-fold cross validation were 0.921 and 0.651 respectively. An interpretability method was applied to explain individual predictions and gain insight into the overall factors contributing to chronic homelessness among the population studied. The model achieves state-of-the-art performance and improved stakeholder trust of what is usually a "black box" neural network model through interpretable AI.
In todays world the request for very complex models for huge data sets is rising steadily. The problem with these models is that by raising the complexity of the models, it gets much harder to interpret them. The growing field of \emph{interpretable machine learning} tries to make up for the lack of interpretability in these complex (or even blackbox-)models by using specific techniques that can help to understand those models better. In this article a new kind of interpretable machine learning method is presented, which can help to understand the partitioning of the feature space into predicted classes in a classification model using quantile shifts. To illustrate in which situations this quantile shift method (QSM) could become beneficial, it is applied to a theoretical medical example and a real data example. Basically, real data points (or specific points of interest) are used and the changes of the prediction after slightly raising or decreasing specific features are observed. By comparing the predictions before and after the manipulations, under certain conditions the observed changes in the predictions can be interpreted as neighborhoods of the classes with regard to the manipulated features. Chordgraphs are used to visualize the observed changes.
Explaining recommendations enables users to understand whether recommended items are relevant to their needs and has been shown to increase their trust in the system. More generally, if designing explainable machine learning models is key to check the sanity and robustness of a decision process and improve their efficiency, it however remains a challenge for complex architectures, especially deep neural networks that are often deemed "black-box". In this paper, we propose a novel formulation of interpretable deep neural networks for the attribution task. Differently to popular post-hoc methods, our approach is interpretable by design. Using masked weights, hidden features can be deeply attributed, split into several input-restricted sub-networks and trained as a boosted mixture of experts. Experimental results on synthetic data and real-world recommendation tasks demonstrate that our method enables to build models achieving close predictive performances to their non-interpretable counterparts, while providing informative attribution interpretations.