Infinite-dimensional next-generation reservoir computing
Authors:
Lyudmila Grigoryeva,
Hannah Lim Jing Ting,
Juan-Pablo Ortega
Abstract:
Next-generation reservoir computing (NG-RC) has attracted much attention due to its excellent performance in spatio-temporal forecasting of complex systems and its ease of implementation. This paper shows that NG-RC can be encoded as a kernel ridge regression that makes training efficient and feasible even when the space of chosen polynomial features is very large. Additionally, an extension to an…
▽ More
Next-generation reservoir computing (NG-RC) has attracted much attention due to its excellent performance in spatio-temporal forecasting of complex systems and its ease of implementation. This paper shows that NG-RC can be encoded as a kernel ridge regression that makes training efficient and feasible even when the space of chosen polynomial features is very large. Additionally, an extension to an infinite number of covariates is possible, which makes the methodology agnostic with respect to the lags into the past that are considered as explanatory factors, as well as with respect to the number of polynomial covariates, an important hyperparameter in traditional NG-RC. We show that this approach has solid theoretical backing and good behavior based on kernel universality properties previously established in the literature. Various numerical illustrations show that these generalizations of NG-RC outperform the traditional approach in several forecasting applications.
△ Less
Submitted 21 February, 2025; v1 submitted 12 December, 2024;
originally announced December 2024.
On the Trust and Trust Modelling for the Future Fully-Connected Digital World: A Comprehensive Study
Authors:
Hannah Lim Jing Ting,
Xin Kang,
Tieyan Li,
Haiguang Wang,
Cheng-Kang Chu
Abstract:
With the fast development of digital technologies, we are running into a digital world. The relationship among people and the connections among things become more and more complex, and new challenges arise. To tackle these challenges, trust-a soft security mechanism-is considered as a promising technology. Thus, in this survey, we do a comprehensive study on the trust and trust modelling for the f…
▽ More
With the fast development of digital technologies, we are running into a digital world. The relationship among people and the connections among things become more and more complex, and new challenges arise. To tackle these challenges, trust-a soft security mechanism-is considered as a promising technology. Thus, in this survey, we do a comprehensive study on the trust and trust modelling for the future digital world. We revisit the definitions and properties of trust, and analysis the trust theories and discuss their impact on digital trust modelling. We analyze the digital world and its corresponding environment where people, things, and infrastructure connect with each other. We detail the challenges that require trust in these digital scenarios. Under our analysis of trust and the digital world, we define different types of trust relationships and find out the factors that are needed to ensure a fully representative model. Next, to meet the challenges of digital trust modelling, comprehensive trust model evaluation criteria are proposed, and potential securities and privacy issues of trust modelling are analyzed. Finally, we provide a wide-ranging analysis of different methodologies, mathematical theories, and how they can be applied to trust modelling.
△ Less
Submitted 14 June, 2021;
originally announced June 2021.
