Locally and globally explainable time series tweaking

Karlsson, Isak; Rebane, Jonathan; Papapetrou, Panagiotis; Gionis, Aristides

doi:10.1007/s10115-019-01389-4

Cited by 22 publications

(12 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Model-agnostic explanation models are typically based on decision trees, rules or feature importance ( Guidotti et al, 2018 ; Freitas, 2014 ; Craven & Shavlik, 1995 ), because of the simplicity of such explanations. Several model-specific and data-specific explanation models have also been developed, e.g., for deep neural networks ( Binder et al, 2016 ; Selvaraju et al, 2019 ), deep relational machines ( Srinivasan, Vig & Bain, 2019 ), time series ( Karlsson et al, 2019 ), multi-labelled and ontology-linked data ( Panigutti, Perotti & Pedreschi, 2020 ) or logic problems ( Biecek, 2018 ); software toolkits including the implementation of various XAI algorithms have been also introduced ( Arya et al, 2019 ). A comprehensive survey of explainability methods can be found in Guidotti et al (2018) and in Došilović, Brčić & Hlupić (2018) .…”

Section: Introductionmentioning

confidence: 99%

To trust or not to trust an explanation: using LEAF to evaluate local linear XAI methods

Amparore

Perotti

Bajardi

2021

PeerJ Computer Science

View full text Add to dashboard Cite

The main objective of eXplainable Artificial Intelligence (XAI) is to provide effective explanations for black-box classifiers. The existing literature lists many desirable properties for explanations to be useful, but there is a scarce consensus on how to quantitatively evaluate explanations in practice. Moreover, explanations are typically used only to inspect black-box models, and the proactive use of explanations as a decision support is generally overlooked. Among the many approaches to XAI, a widely adopted paradigm is Local Linear Explanations—with LIME and SHAP emerging as state-of-the-art methods. We show that these methods are plagued by many defects including unstable explanations, divergence of actual implementations from the promised theoretical properties, and explanations for the wrong label. This highlights the need to have standard and unbiased evaluation procedures for Local Linear Explanations in the XAI field. In this paper we address the problem of identifying a clear and unambiguous set of metrics for the evaluation of Local Linear Explanations. This set includes both existing and novel metrics defined specifically for this class of explanations. All metrics have been included in an open Python framework, named LEAF. The purpose of LEAF is to provide a reference for end users to evaluate explanations in a standardised and unbiased way, and to guide researchers towards developing improved explainable techniques.

show abstract

Section: Introductionmentioning

confidence: 99%

To trust or not to trust an explanation: using LEAF to evaluate local linear XAI methods

Amparore

Perotti

Bajardi

2021

PeerJ Computer Science

View full text Add to dashboard Cite

show abstract

“…The created simple model regularly seeks to optimize its similarity to the original function while minimizing the complexity and achieving comparable performance to the original model. There are two methods that can be considered subsets of the simplification approach, which include the local explanation (Neto and Paulovich 2020;Karlsson et al 2020) and the example generation (Chen et al 2021). The local explanation assumes that simplification can be achieved by segmenting the solution space into smaller segments and performing an analysis on the particular segments of a model.…”

Section: Simplificationmentioning

confidence: 99%

Explainable artificial intelligence: a comprehensive review

et al. 2021

View full text Add to dashboard Cite

Thanks to the exponential growth in computing power and vast amounts of data, artificial intelligence (AI) has witnessed remarkable developments in recent years, enabling it to be ubiquitously adopted in our daily lives. Even though AI-powered systems have brought competitive advantages, the black-box nature makes them lack transparency and prevents them from explaining their decisions. This issue has motivated the introduction of explainable artificial intelligence (XAI), which promotes AI algorithms that can show their internal process and explain how they made decisions. The number of XAI research has increased significantly in recent years, but there lacks a unified and comprehensive review of the latest XAI progress. This review aims to bridge the gap by discovering the critical perspectives of the rapidly growing body of research associated with XAI. After offering the readers a solid XAI background, we analyze and review various XAI methods, which are grouped into (i) pre-modeling explainability, (ii) interpretable model, and (iii) post-modeling explainability. We also pay attention to the current methods that dedicate to interpret and analyze deep learning methods. In addition, we systematically discuss various XAI challenges, such as the trade-off between the performance and the explainability, evaluation methods, security, and policy. Finally, we show the standard approaches that are leveraged to deal with the mentioned challenges.

show abstract

“…Moreover, instance-or example-based explanations can be more easily interpreted by a non-expert person [31]. These methods explain a prediction on a single instance by comparing it to another real or generated example, e.g., the most typical examplar of the observed phenomenon (a prototype [16]) or a contrastive examplar related to a distinct behaviour (a counterfactual [2, 14,18]). For time series classifiers, counterfactuals can be generated by swapping the values of the most discriminative dimensions with those from another training instance [2].…”

Section: Related Workmentioning

confidence: 99%

“…The Native Guide algorithm [14] does not suffer from the previous issue but uses a perturbation mechanism on the Nearest Unlike Neighbor in the training set using the model's internal feature vector. Lastly, for a k-NN and a Random Shapelet Forest classifiers, [18] design a tweaking mechanism to produce counterfactual time series.…”

Section: Related Workmentioning

confidence: 99%

“…With these properties in mind, we aim to generate counterfactual samples for explaining anomalies. Counterfactual explanations have been previously proposed for time series classifiers [2, 14,18] and can achieve criterion (a). A counterfactual example (or for short, a counterfactual) is an instance-based explanation in the form of a perturbed input on which the model's prediction value is different from the original model output.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Diverse Counterfactual Explanations for Anomaly Detection in Time Series

Sulem¹,

Donini²,

Zafar³

et al. 2022

Preprint

View full text Add to dashboard Cite

Data-driven methods that detect anomalies in times series data are ubiquitous in practice, but they are in general unable to provide helpful explanations for the predictions they make. In this work we propose a model-agnostic algorithm that generates counterfactual ensemble explanations for time series anomaly detection models. Our method generates a set of diverse counterfactual examples, i.e, multiple perturbed versions of the original time series that are not considered anomalous by the detection model. Since the magnitude of the perturbations is limited, these counterfactuals represent an ensemble of inputs similar to the original time series that the model would deem normal. Our algorithm is applicable to any differentiable anomaly detection model. We investigate the value of our method on univariate and multivariate real-world datasets and two deep-learning-based anomaly detection models, under several explainability criteria previously proposed in other data domains such as Validity, Plausibility, Closeness and Diversity. We show that our algorithm can produce ensembles of counterfactual examples that satisfy these criteria and thanks to a novel type of visualization, can convey a richer interpretation of a model's internal mechanism than existing methods. Moreover, we design a sparse variant of our method to improve the interpretability of counterfactual explanations for high-dimensional time series anomalies. In this setting, our explanation is localized on only a few dimensions and can therefore be communicated more efficiently to the model's user.

show abstract

Locally and globally explainable time series tweaking

Cited by 22 publications

References 38 publications

To trust or not to trust an explanation: using LEAF to evaluate local linear XAI methods

To trust or not to trust an explanation: using LEAF to evaluate local linear XAI methods

Explainable artificial intelligence: a comprehensive review

Diverse Counterfactual Explanations for Anomaly Detection in Time Series

Contact Info

Product

Resources

About