DeepAI AI Chat
Log In Sign Up

Universal Model-free Information Extraction

by   Bin Li, et al.

Bayesian approaches have been used extensively in scientific and engineering research to quantify uncertainty and extract information. However, its model-dependent nature means that when the a priori model is incomplete or unavailable, there is a severe risk that Bayesian approaches will yield misleading results. Here, we propose a universal model-free information extraction approach, capable of reliably recovering target signals from complex responses. This breakthrough leverages on a data-centric approach, whereby measured data is reconfigured to create an enriched observable space, which in turn is mapped to a well-adapted manifold, thereby detecting crucial information via a reconstructed low-rank phase-space. A Koopman operator is used to transform hidden and complex nonlinear dynamics to linear one, which enables us to detect hidden event of interest from rapidly evolving systems, and relate it to either unobservable stimulus or anomalous behaviour. Thanks to its data-driven nature, our method excludes completely any prior knowledge on governing dynamics. We benchmark the astonishing accuracy of our method on three diverse and challenging problems in: biology, medicine, and engineering. In all cases, our approach outperforms existing state-of-the-art methods, of both Bayesian and non-Bayesian type. By creating a new reliable information analysis paradigm, it is suitable for ubiquitous nonlinear dynamical systems or end-users with little expertise, which permits the unbiased understanding of various mechanisms in the real world.


page 1

page 2

page 3

page 4


Data Driven Control with Learned Dynamics: Model-Based versus Model-Free Approach

This paper compares two different types of data-driven control methods, ...

Deep Learning of Koopman Representation for Control

We develop a data-driven, model-free approach for the optimal control of...

Stabilizing Dynamical Systems via Policy Gradient Methods

Stabilizing an unknown control system is one of the most fundamental pro...

Universal set of Observables for the Koopman Operator through Causal Embedding

Obtaining repeated measurements from physical and natural systems for bu...

Sampling and Inference of Networked Dynamics using Log-Koopman Nonlinear Graph Fourier Transform

Networked nonlinear dynamics underpin the complex functionality of many ...

Bayesian Nonlinear Models for Repeated Measurement Data: An Overview, Implementation, and Applications

Nonlinear mixed effects models have become a standard platform for analy...