Modern predictive analytics underpinned by machine learning techniques has become a key enabler to the automation of data-driven decision making. In the context of business process management, predictive analytics has been applied to making predictions about the future state of an ongoing business process instance, for example, when will the process instance complete and what will be the outcome upon completion. Machine learning models can be trained on event log data recording historical process execution to build the underlying predictive models. Multiple techniques have been proposed so far which encode the information available in an event log and construct input features required to train a predictive model. While accuracy has been a dominant criterion in the choice of various techniques, they are often applied as a black-box in building predictive models. In this paper, we derive explanations using interpretable machine learning techniques to compare and contrast the suitability of multiple predictive models of high accuracy. The explanations allow us to gain an understanding of the underlying reasons for a prediction and highlight scenarios where accuracy alone may not be sufficient in assessing the suitability of techniques used to encode event log data to features used by a predictive model. Findings from this study motivate the need and importance to incorporate interpretability in predictive process analytics.
Multi-horizon forecasting problems often contain a complex mix of inputs -- including static (i.e. time-invariant) covariates, known future inputs, and other exogenous time series that are only observed historically -- without any prior information on how they interact with the target. While several deep learning models have been proposed for multi-step prediction, they typically comprise black-box models which do not account for the full range of inputs present in common scenarios. In this paper, we introduce the Temporal Fusion Transformer (TFT) -- a novel attention-based architecture which combines high-performance multi-horizon forecasting with interpretable insights into temporal dynamics. To learn temporal relationships at different scales, the TFT utilizes recurrent layers for local processing and interpretable self-attention layers for learning long-term dependencies. The TFT also uses specialized components for the judicious selection of relevant features and a series of gating layers to suppress unnecessary components, enabling high performance in a wide range of regimes. On a variety of real-world datasets, we demonstrate significant performance improvements over existing benchmarks, and showcase three practical interpretability use-cases of TFT.
Glioblastoma multiforme (GBM) is a fast-growing and highly invasive brain tumour, it tends to occur in adults between the ages of 45 and 70 and it accounts for 52 percent of all primary brain tumours. Usually, GBMs are detected by magnetic resonance images (MRI). Among MRI, Fluid-attenuated inversion recovery (FLAIR) sequence produces high quality digital tumour representation. Fast detection and segmentation techniques are needed for overcoming subjective medical doctors (MDs) judgment. In the present investigation, we intend to demonstrate by means of numerical experiments that topological features combined with textural features can be enrolled for GBM analysis and morphological characterization on FLAIR. To this extent, we have performed three numerical experiments. In the first experiment, Topological Data Analysis (TDA) of a simplified 2D tumour growth mathematical model had allowed to understand the bio-chemical conditions that facilitate tumour growth: the higher the concentration of chemical nutrients the more virulent the process. In the second experiment topological data analysis was used for evaluating GBM temporal progression on FLAIR recorded within 90 days following treatment (e.g., chemo-radiation therapy - CRT) completion and at progression. The experiment had confirmed that persistent entropy is a viable statistics for monitoring GBM evolution during the follow-up period. In the third experiment we had developed a novel methodology based on topological and textural features and automatic interpretable machine learning for automatic GBM classification on FLAIR. The algorithm reached a classification accuracy up to the 97%.
A salient approach to interpretable machine learning is to restrict modeling to simple models. In the Bayesian framework, this can be pursued by restricting the model structure and prior to favor interpretable models. Fundamentally, however, interpretability is about users' preferences, not the data generation mechanism; it is more natural to formulate interpretability as a utility function. In this work, we propose an interpretability utility, which explicates the trade-off between explanation fidelity and interpretability in the Bayesian framework. The method consists of two steps. First, a reference model, possibly a black-box Bayesian predictive model which does not compromise accuracy, is fitted to the training data. Second, a proxy model from an interpretable model family that best mimics the predictive behaviour of the reference model is found by optimizing the interpretability utility function. The approach is model agnostic -- neither the interpretable model nor the reference model are restricted to a certain class of models -- and the optimization problem can be solved using standard tools. Through experiments on real-word data sets, using decision trees as interpretable models and Bayesian additive regression models as reference models, we show that for the same level of interpretability, our approach generates more accurate models than the alternative of restricting the prior. We also propose a systematic way to measure stability of interpretabile models constructed by different interpretability approaches and show that our proposed approach generates more stable models.
The problem of attribution is concerned with identifying the parts of an input that are responsible for a model's output. An important family of attribution methods is based on measuring the effect of perturbations applied to the input. In this paper, we discuss some of the shortcomings of existing approaches to perturbation analysis and address them by introducing the concept of extremal perturbations, which are theoretically grounded and interpretable. We also introduce a number of technical innovations to compute extremal perturbations, including a new area constraint and a parametric family of smooth perturbations, which allow us to remove all tunable hyper-parameters from the optimization problem. We analyze the effect of perturbations as a function of their area, demonstrating excellent sensitivity to the spatial properties of the deep neural network under stimulation. We also extend perturbation analysis to the intermediate layers of a network. This application allows us to identify the salient channels necessary for classification, which, when visualized using feature inversion, can be used to elucidate model behavior. Lastly, we introduce TorchRay, an interpretability library built on PyTorch.
Deep learning has enjoyed much recent success, and applying state-of-the-art model learning methods to controls is an exciting prospect. However, there is a strong reluctance to use these methods on safety-critical systems, which have constraints on safety, stability, and real-time performance. We propose a framework which satisfies these constraints while allowing the use of deep neural networks for learning model uncertainties. Central to our method is the use of Bayesian model learning, which provides an avenue for maintaining appropriate degrees of caution in the face of the unknown. In the proposed approach, we develop an adaptive control framework leveraging the theory of stochastic CLFs (Control Lyapunov Functions) and stochastic CBFs (Control Barrier Functions) along with tractable Bayesian model learning via Gaussian Processes or Bayesian neural networks. Under reasonable assumptions, we guarantee stability and safety while adapting to unknown dynamics with probability 1. We demonstrate this architecture for high-speed terrestrial mobility targeting potential applications in safety-critical high-speed Mars rover missions.
Being able to interpret, or explain, the predictions made by a machine learning model is of fundamental importance. This is especially true when there is interest in deploying data-driven models to make high-stakes decisions, e.g. in healthcare. While recent years have seen an increasing interest in interpretable machine learning research, this field is currently lacking an agreed-upon definition of interpretability, and some researchers have called for a more active conversation towards a rigorous approach to interpretability. Joining this conversation, we claim in this paper that the difficulty of interpreting a complex model stems from the existing interactions among features. We argue that by enforcing monotonicity between features and outputs, we are able to reason about the effect of a single feature on an output independently from other features, and consequently better understand the model. We show how to structurally introduce this constraint in deep learning models by adding new simple layers. We validate our model on benchmark datasets, and compare our results with previously proposed interpretable models.
With a mortality rate of 5.4 million lives worldwide every year and a healthcare cost of more than 16 billion dollars in the USA alone, sepsis is one of the leading causes of hospital mortality and an increasing concern in the ageing western world. Recently, medical and technological advances have helped re-define the illness criteria of this disease, which is otherwise poorly understood by the medical society. Together with the rise of widely accessible Electronic Health Records, the advances in data mining and complex nonlinear algorithms are a promising avenue for the early detection of sepsis. This work contributes to the research effort in the field of automated sepsis detection with an open-access labelling of the medical MIMIC-III data set. Moreover, we propose MGP-AttTCN: a joint multitask Gaussian Process and attention-based deep learning model to early predict the occurrence of sepsis in an interpretable manner. We show that our model outperforms the current state-of-the-art and present evidence that different labelling heuristics lead to discrepancies in task difficulty. For instance, when predicting sepsis five hours prior to onset on our new realistic labels, our proposed model achieves an area under the ROC curve of 0.660 and an area under the PR curve of 0.483, whereas the (less interpretable) previous state-of-the-art model (MGP-TCN) achieves 0.635 AUROC and 0.460 AUPR and the popular commercial InSight model achieves 0.490 AUROC and 0.359 AUPR.
Purpose: Manual feedback from senior surgeons observing less experienced trainees is a laborious task that is very expensive, time-consuming and prone to subjectivity. With the number of surgical procedures increasing annually, there is an unprecedented need to provide an accurate, objective and automatic evaluation of trainees' surgical skills in order to improve surgical practice. Methods: In this paper, we designed a convolutional neural network (CNN) to classify surgical skills by extracting latent patterns in the trainees' motions performed during robotic surgery. The method is validated on the JIGSAWS dataset for two surgical skills evaluation tasks: classification and regression. Results: Our results show that deep neural networks constitute robust machine learning models that are able to reach new competitive state-of-the-art performance on the JIGSAWS dataset. While we leveraged from CNNs' efficiency, we were able to minimize its black-box effect using the class activation map technique. Conclusions: This characteristic allowed our method to automatically pinpoint which parts of the surgery influenced the skill evaluation the most, thus allowing us to explain a surgical skill classification and provide surgeons with a novel personalized feedback technique. We believe this type of interpretable machine learning model could integrate within "Operation Room 2.0" and support novice surgeons in improving their skills to eventually become experts.
Among interpretable machine learning methods, the class of Generalised Additive Neural Networks (GANNs) is referred to as Self-Explaining Neural Networks (SENN) because of the linear dependence on explicit functions of the inputs. In binary classification this shows the precise weight that each input contributes towards the logit. The nomogram is a graphical representation of these weights. We show that functions of individual and pairs of variables can be derived from a functional Analysis of Variance (ANOVA) representation, enabling an efficient feature selection to be carried by application of the logistic Lasso. This process infers the structure of GANNs which otherwise needs to be predefined. As this method is particularly suited for tabular data, it starts by fitting a generic flexible model, in this case a Multi-layer Perceptron (MLP) to which the ANOVA decomposition is applied. This has the further advantage that the resulting GANN can be replicated as a SENN, enabling further refinement of the univariate and bivariate component functions to take place. The component functions are partial responses hence the SENN is a partial response network. The Partial Response Network (PRN) is equally as transparent as a traditional logistic regression model, but capable of non-linear classification with comparable or superior performance to the original MLP. In other words, the PRN is a fully interpretable representation of the MLP, at the level of univariate and bivariate effects. The performance of the PRN is shown to be competitive for benchmark data, against state-of-the-art machine learning methods including GBM, SVM and Random Forests. It is also compared with spline-based Sparse Additive Models (SAM) showing that a semi-parametric representation of the GAM as a neural network can be as effective as the SAM though less constrained by the need to set spline nodes.
There is an increasing interest in estimating heterogeneity in causal effects in randomized and observational studies. However, little research has been conducted to understand heterogeneity in an instrumental variables study. In this work, we present a method to estimate heterogeneous causal effects using an instrumental variable approach. The method has two parts. The first part uses subject-matter knowledge and interpretable machine learning techniques, such as classification and regression trees, to discover potential effect modifiers. The second part uses closed testing to test for the statistical significance of the effect modifiers while strongly controlling familywise error rate. We conducted this method on the Oregon Health Insurance Experiment, estimating the effect of Medicaid on the number of days an individual's health does not impede their usual activities, and found evidence of heterogeneity in older men who prefer English and don't self-identify as Asian and younger individuals who have at most a high school diploma or GED and prefer English.
Takagi-Sugeno-Kang (TSK) fuzzy systems are flexible and interpretable machine learning models; however, they may not be easily optimized when the data size is large, and/or the data dimensionality is high. This paper proposes a mini-batch gradient descent (MBGD) based algorithm to efficiently and effectively train TSK fuzzy classifiers. It integrates two novel techniques: 1) uniform regularization (UR), which forces the rules to have similar average contributions to the output, and hence to increase the generalization performance of the TSK classifier; and, 2) batch normalization (BN), which extends BN from deep neural networks to TSK fuzzy classifiers to expedite the convergence and improve the generalization performance. Experiments on 12 UCI datasets from various application domains, with varying size and dimensionality, demonstrated that UR and BN are effective individually, and integrating them can further improve the classification performance.
Interpretable Machine Learning (IML) has become increasingly important in many real-world applications, such as autonomous cars and medical diagnosis, where explanations are significantly preferred to help people better understand how machine learning systems work and further enhance their trust towards systems. However, due to the diversified scenarios and subjective nature of explanations, we rarely have the ground truth for benchmark evaluation in IML on the quality of generated explanations. Having a sense of explanation quality not only matters for assessing system boundaries, but also helps to realize the true benefits to human users in practical settings. To benchmark the evaluation in IML, in this article, we rigorously define the problem of evaluating explanations, and systematically review the existing efforts from state-of-the-arts. Specifically, we summarize three general aspects of explanation (i.e., generalizability, fidelity and persuasibility) with formal definitions, and respectively review the representative methodologies for each of them under different tasks. Further, a unified evaluation framework is designed according to the hierarchical needs from developers and end-users, which could be easily adopted for different scenarios in practice. In the end, open problems are discussed, and several limitations of current evaluation techniques are raised for future explorations.
Recent research has demonstrated that feature attribution methods for deep networks can themselves be incorporated into training; these attribution priors optimize for a model whose attributions have certain desirable properties -- most frequently, that particular features are important or unimportant. These attribution priors are often based on attribution methods that are not guaranteed to satisfy desirable interpretability axioms, such as completeness and implementation invariance. Here, we introduce attribution priors to optimize for higher-level properties of explanations, such as smoothness and sparsity, enabled by a fast new attribution method formulation called expected gradients that satisfies many important interpretability axioms. This improves model performance on many real-world tasks where previous attribution priors fail. Our experiments show that the gains from combining higher-level attribution priors with expected gradients attributions are consistent across image, gene expression, and health care data sets. We believe this work motivates and provides the necessary tools to support the widespread adoption of axiomatic attribution priors in many areas of applied machine learning. The implementations and our results have been made freely available to academic communities.
The interpretability of machine learning, particularly for deep neural networks, is crucial for decision making in real-world applications. One approach is replacing the un-interpretable machine learning model with a surrogate model, which has a simple structure for interpretation. Another approach is understanding the target system by using a simulation modeled by human knowledge with interpretable simulation parameters. Recently, simulator calibration has been developed based on kernel mean embedding to estimate the simulation parameters as posterior distributions. Our idea is to use a simulation model as an interpretable surrogate model. However, the computational cost of simulator calibration is high owing to the complexity of the simulation model. Thus, we propose a ''model-bridging'' framework to bridge machine learning models with simulation models by a series of kernel mean embeddings to address these difficulties. The proposed framework enables us to obtain predictions and interpretable simulation parameters simultaneously without the computationally expensive calculations of the simulations. In this study, we apply the proposed framework to essential simulations in the manufacturing industry, such as production simulation and fluid dynamics simulation.
Explainability in Artificial Intelligence has been revived as a topic of active research by the need of conveying safety and trust to users in the `how' and `why' of automated decision-making. Whilst a plethora of approaches have been developed for post-hoc explainability, only a few focus on how to use domain knowledge, and how this influences the understandability of global explanations from the users' perspective. In this paper, we show how ontologies help the understandability of global post-hoc explanations, presented in the form of symbolic models. In particular, we build on Trepan, an algorithm that explains artificial neural networks by means of decision trees, and we extend it to include ontologies modeling domain knowledge in the process of generating explanations. We present the results of a user study that measures the understandability of decision trees using a syntactic complexity measure, and through time and accuracy of responses as well as reported user confidence and understandability. The user study considers domains where explanations are critical, namely, in finance and medicine. The results show that decision trees generated with our algorithm, taking into account domain knowledge, are more understandable than those generated by standard Trepan without the use of ontologies.
For sensitive problems, such as medical imaging or fraud detection, Neural Network (NN) adoption has been slow due to concerns about their reliability, leading to a number of algorithms for explaining their decisions. NNs have also been found vulnerable to a class of imperceptible attacks, called adversarial examples, which arbitrarily alter the output of the network. Here we demonstrate both that these attacks can invalidate prior attempts to explain the decisions of NNs, and that with very robust networks, the attacks themselves may be leveraged as explanations with greater fidelity to the model. We show that the introduction of a novel regularization technique inspired by the Lipschitz constraint, alongside other proposed improvements, greatly improves an NN's resistance to adversarial examples. On the ImageNet classification task, we demonstrate a network with an Accuracy-Robustness Area (ARA) of 0.0053, an ARA 2.4x greater than the previous state of the art. Improving the mechanisms by which NN decisions are understood is an important direction for both establishing trust in sensitive domains and learning more about the stimuli to which NNs respond.
Most of the work on interpretable machine learning has focused on designing either inherently interpretable models, which typically trade-off accuracy for interpretability, or post-hoc explanation systems, which lack guarantees about their explanation quality. We propose an alternative to these approaches by directly regularizing a black-box model for interpretability at training time. Our approach explicitly connects three key aspects of interpretable machine learning: (i) the model's innate explainability, (ii) the explanation system used at test time, and (iii) the metrics that measure explanation quality. Our regularization results in substantial improvement in terms of the explanation fidelity and stability metrics across a range of datasets and black-box explanation systems while slightly improving accuracy. Further, if the resulting model is still not sufficiently interpretable, the weight of the regularization term can be adjusted to achieve the desired trade-off between accuracy and interpretability. Finally, we justify theoretically that the benefits of explanation-based regularization generalize to unseen points.
Machine learning has recently been widely adopted to address the managerial decision making problems, in which the decision maker needs to be able to interpret the contributions of individual attributes in an explicit form. However, there is a trade-off between performance and interpretability. Full complexity models are non-traceable black-box, whereas classic interpretable models are usually simplified with lower accuracy. This trade-off limits the application of state-of-the-art machine learning models in management problems, which requires high prediction performance, as well as the understanding of individual attributes' contributions to the model outcome. Multiple criteria decision aiding (MCDA) is a family of analytic approaches to depicting the rationale of human decision. It is also limited by strong assumptions. To meet the decision maker's demand for more interpretable machine learning models, we propose a novel hybrid method, namely Neural Network-based Multiple Criteria Decision Aiding, which combines an additive value model and a fully-connected multilayer perceptron (MLP) to achieve good performance while capturing the explicit relationships between individual attributes and the prediction. NN-MCDA has a linear component to characterize such relationships through providing explicit marginal value functions, and a nonlinear component to capture the implicit high-order interactions between attributes and their complex nonlinear transformations. We demonstrate the effectiveness of NN-MCDA with extensive simulation studies and three real-world datasets. To the best of our knowledge, this research is the first to enhance the interpretability of machine learning models with MCDA techniques. The proposed framework also sheds light on how to use machine learning techniques to free MCDA from strong assumptions.
Tree ensembles, such as random forests and AdaBoost, are ubiquitous machine learning models known for achieving strong predictive performance across a wide variety of domains. However, this strong performance comes at the cost of interpretability (i.e. users are unable to understand the relationships a trained random forest has learned and why it is making its predictions). In particular, it is challenging to understand how the contribution of a particular feature, or group of features, varies as their value changes. To address this, we introduce Disentangled Attribution Curves (DAC), a method to provide interpretations of tree ensemble methods in the form of (multivariate) feature importance curves. For a given variable, or group of variables, DAC plots the importance of a variable(s) as their value changes. We validate DAC on real data by showing that the curves can be used to increase the accuracy of logistic regression while maintaining interpretability, by including DAC as an additional feature. In simulation studies, DAC is shown to out-perform competing methods in the recovery of conditional expectations. Finally, through a case-study on the bike-sharing dataset, we demonstrate the use of DAC to uncover novel insights into a dataset.