Parameter estimation in dynamical systems via Statistical Learning: a reinterpretation of Approximate Bayesian Computation applied to COVID-19 spread
We propose a robust parameter estimation method for dynamical systems based on Statistical Learning techniques which aims to estimate a set of parameters that well fit the dynamics in order to obtain robust evidences about the qualitative behaviour of its trajectory. The method is quite general and flexible, since it does not rely on any specific property of the dynamical system, and represents a reinterpretation of Approximate Bayesian Computation methods through the lens of Statistical Learning. The method is specially useful for estimating parameters in epidemiological compartmental models in order to obtain qualitative properties of a disease evolution. We apply it to simulated and real data about COVID-19 spread in the US in order to evaluate qualitatively its evolution over time, showing how one may assess the effectiveness of measures implemented to slow the spread and some qualitative features of the disease current and future evolution.
Code (0)
등록된 구현이 없습니다.
Tasks
parameter estimationSimilar Papers 제목 키워드 기반
Multi-view Matrix Factorization for Linear Dynamical System Estimation
We consider maximum likelihood estimation of linear dynamical systems with generalized-linear observation models. Maximum likelihood is typically considered to be hard in this setting since latent states and transition p…
global-optimizationDeep Learning of Dynamical System Parameters from Return Maps as Images
We present a novel approach to system identification (SI) using deep learning techniques. Focusing on parametric system identification (PSI), we use a supervised learning approach for estimating the parameters of discret…
Data Augmentationparameter estimationOn the Nonasymptotic Scaling Guarantee of Hyperparameter Estimation in Inhomogeneous, Weakly-Dependent Complex Network Dynamical Systems
Hierarchical Bayesian models are increasingly used in large, inhomogeneous complex network dynamical systems by modeling parameters as draws from a hyperparameter-governed distribution. However, theoretical guarantees fo…
Second-order AAA algorithms for structured data-driven modeling
The data-driven modeling of dynamical systems has become an essential tool for the construction of accurate computational models from real-world data. In this process, the inherent differential structures underlying the …
Transfer Learning for Neural Parameter Estimation applied to Building RC Models
Parameter estimation for dynamical systems remains challenging due to non-convexity and sensitivity to initial parameter guesses. Recent deep learning approaches enable accurate and fast parameter estimation but do not e…
Transfer Learning