Explaining Memristive Reservoir Computing Through Evolving Feature Attribution

Shi，Xinming1; Wang，Zilu2; Minku，Leandro L.3; Yao，Xin1

10.1145/3583133.3590619

2023

Genetic and Evolutionary Computation Conference (GECCO)

PROCEEDINGS OF THE 2023 GENETIC AND EVOLUTIONARY COMPUTATION CONFERENCE COMPANION, GECCO 2023 COMPANION

JUL 15-19, 2023

null,Lisbon,PORTUGAL

1601 Broadway, 10th Floor, NEW YORK, NY, UNITED STATES

ASSOC COMPUTING MACHINERY

Memristive Reservoir Computing (MRC) is a promising computing architecture for time series tasks, but lacks explainability, leading to unreliable predictions. To address this issue, we propose an evolutionary framework to explain the time series predictions of MRC systems. Our proposed approach attributes the feature importance of the time series via an evolutionary approach to explain the predictions. Our experiments show that our approach successfully identified the most influential factors, demonstrating the effectiveness of our design and its superiority in terms of explanation compared to state-of-the-art methods.

Explainability evolutionary algorithm reservoir computing memristor

第一 ; 通讯

英语

EI ; CPCI-S

NSFC[62250710682]

Computer Science

Computer Science, Artificial Intelligence ; Computer Science, Information Systems

WOS:001117972600193

Web of Science

1.Southern University of Science and Technology,China University of Birmingham,United Kingdom
2.Southern University of Science and Technology,China
3.University of Birmingham,United Kingdom

Shi，Xinming,Wang，Zilu,Minku，Leandro L.,et al. Explaining Memristive Reservoir Computing Through Evolving Feature Attribution[C]. 1601 Broadway, 10th Floor, NEW YORK, NY, UNITED STATES:ASSOC COMPUTING MACHINERY,2023.