timeXplain - A Framework for Explaining the Predictions of Time Series Classifiers (2020-07-15T00:00:00.000000Z)

TL;DR

The timeXplain framework is employed in a large-scale experimental comparison of several state-of-the-art time series classifiers and similarities between seemingly distinct classification concepts such as residual neural networks and elastic ensembles are discovered.

Abstract

Modern time series classifiers display impressive predictive capabilities, yet their decision-making processes mostly remain black boxes to the user. At the same time, model-agnostic explainers, such as the recently proposed SHAP, promise to make the predictions of machine learning models interpretable, provided there are well-designed domain mappings. We bring both worlds together in our timeXplain framework, extending the reach of explainable artificial intelligence to time series classification and value prediction. We present novel domain mappings for the time and the frequency domain as well as series statistics and analyze their explicative power as well as their limits. We employ timeXplain in a large-scale experimental comparison of several state-of-the-art time series classifiers and discover similarities between seemingly distinct classification concepts such as residual neural networks and elastic ensembles.

Authors

Patrick Schäfer

4 papers

Felix Mujkanovic

1 papers

Vanja Doskoč

1 papers

TL;DR

Abstract

Authors

References35 items

From local explanations to global understanding with explainable AI for trees

Agnostic Local Explanation for Time Series Classification

sktime: A Unified Interface for Machine Learning with Time Series

Time Series Prediction Algorithm for Intelligent Predictive Maintenance

Interpretable time series classification using linear models and multi-resolution multi-domain symbolic representations

TIME SERIES

Deep learning for time series classification: a review

Interpretable Time Series Classification using All-Subsequence Learning and Symbolic Representations in Time and Frequency Domains

Techniques for interpretable machine learning

Time Series Classification with HIVE-COTE

On the Robustness of Interpretability Methods

Towards Robust Interpretability with Self-Explaining Neural Networks

Scalable and accurate deep learning with electronic health records

A Unified Approach to Interpreting Model Predictions

Fast and Accurate Time Series Classification with WEASEL

The great time series classification bake off: a review and experimental evaluation of recent algorithmic advances

Scalable time series classification

European Union Regulations on Algorithmic Decision-Making and a "Right to Explanation"

The mythos of model interpretability

“Why Should I Trust You?”: Explaining the Predictions of Any Classifier

Classification of time series by shapelet transformation

Intelligible models for classification and regression

A shapelet transform for time series classification

Scikit-learn: Machine Learning in Python

Real-valued fast Fourier transform algorithms

Dynamic programming algorithm optimization for spoken word recognition

pyts: A Python Package for Time Series Classification

Tslearn, A Machine Learning Toolkit for Time Series Data

A Model-Agnostic Approach to Quantifying the Informativeness of Explanation Methods for Time Series Classification

Time-Series Classification Methods: Review and Applications to Power Systems Data

The UCR time series classification archive (October 2018), https: //www.cs.ucr.edu/~eamonn/time_series_data_2018

SIGNAL-MATCHED POWER-COMPLEMENTARY CROSS-FADING AND DRY-WET MIXING

SAX-VSM: Inter-pretable Time Series Classification Using SAX and Vector Space Model

Discrete Time Signal Processing

Shapelet techniques classify series based on the occurrence of class-speciﬁc, phase-independent subseries found during training, named shapelets

Field of Study

Journal Information

Name

Volume

Venue Information

Name

Type

URL

Alternate Names