TempXAI@ECMLPKDD

TempXAI:
Explainable AI for Time Series and Data Streams Tutorial-Workshop

TempXAI Workshop at the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD) 2024

WORKSHOP Overview

KEYNOTE SPEAKER

Organizers

Program Committee (Preliminary)

Program

VENUE (September 9-13)

WORKSHOP Overview

This full day workshop at ECML PKDD 2024 is an extension of last year's edition of the Explainable AI (XAI) for Time Series and Explainable Artificial Intelligence from Static to Dynamic workshops to a full-day combined joint workshop and tutorial on Explainable AI for Time series and Data Streams.

The workshop focuses on exploring the crucial intersection of Explainable AI (XAI) and the challenges posed by time series and data streams. Our primary objectives include understanding Dynamic Interpretability, delving into techniques that offer transparent insights into time-evolving data, and providing a better understanding of machine learning models in dynamic environments. We aim to advance Incremental Explainability by investigating methods that ensure interpretability remains effective as models adapt to changing data over time or methods that are able to explain these changes. Moreover, we seek to promote Real-time Decision-making by exploring applications of XAI in real-time decision-making scenarios, addressing the need for interpretable models in time-sensitive contexts. The workshop also aims to share practical insights by encouraging the sharing of novel XAI tools that are specific to time series and data streams, in addition to case studies and practical implementations in employing interpretable machine learning for time series and data streams. The TempXAI workshop welcomes papers that cover, but are not limited to, one or several of the following topics:

Explainable AI methods for time series modeling
Explainable AI methods for data streams and models in flux
Interpretable machine learning algorithms for time series and data streams
Explainable deep learning for time series and data stream modeling
Explainable concept drift detection in time series and data streams
Explainable anomaly detection in time series or data streams
Explainable pattern discovery and recognition in time series
Explainability methods for multivariate time series
Explainable time series features engineering
Explainable aggregation of time series
Integration of domain knowledge in time series modeling
Explainability for continual learning and domain adaptation
Visual explanations for (long) temporal data
Causality; Stochastic process modeling
Explainability metrics and evaluation, including benchmark time series and streaming datasets
Case studies and applications of explainable artificial intelligence for time series or data streams
Regulatory compliance and ethics

KEYNOTE SPEAKER

Riccardo Guidotti

University of Pisa

Explanation Methods for Sequential Data Model - From Post-hoc to Interpretable-by-design Approaches for Time Series Classification

The increasing availability of high-dimensional time series data, such as electrocardiograms, stock indices, and motion sensors, has led to the widespread use of time series classifiers in critical fields like healthcare, finance, and transportation. However, the complexity of these models often makes them black boxes, hindering interpretability. In high-stakes domains, explaining a model’s decisions is vital for trust and accountability. Effective eXplainable AI (XAI) methods for sequential data are essential for providing insights and reinforcing expert decision-making. This presentation addresses the challenge of explaining sequential data models, focusing on time series classification. We begin by reviewing the current literature on XAI for time series classification. Then, we present a series of works that illustrate the transition from general-purpose post-hoc explanation approaches to interpretable-by-design methods. First, we introduce a local post-hoc agnostic subsequence-based time series explainer that can be used to elucidate the predictions of any time series classifier. Next, we demonstrate, through a real case study on car crash prediction, how insights from a post-hoc explainer were crucial in developing an effective interpretable-by-design method. Additionally, we showcase an interpretable subsequence-based classifier by enhancing SAX with dilation and stride to capture temporal patterns effectively. Finally, we explore the use of subsequence-based approaches in other sequential domains like mobility trajectories and text.