Stochastic Approach for System Identification using Machine Learning

Estimation of functions from sparse and noisy data is a central theme in machine learning. In the last few years, many algorithms have been developed that exploit Tikhonov regularization theory and reproducing kernel Hilbert spaces. These are the so-called kernel-based methods, which include powerful approaches like regularization networks, support vector machines, and Gaussian regression. Recently, these techniques have also gained popularity in the system identification community. In both linear and nonlinear settings, kernels that incorporate information on dynamic systems, such as the smoothness and stability of the input–output map, can challenge consolidated approaches based on parametric model structures. In the classical parametric setting, the complexity of the model (the model order) needs to be chosen, typically from a finite family of alternatives, by trading bias and variance. This (discrete) model order selection step may be critical, especially when the true model does not belong to the model class. In regularization-based approaches, model complexity is controlled by tuning (continuous) regularization parameters, making the model selection step more robust. In this article, we review these new kernel-based system identification approaches and discuss extensions based on nuclear and [Formula: see text] norms.

Download Full-text

System identification models for waves in the inner magnetosphere

10.5194/egusphere-egu2020-16544 ◽

2020 ◽

Author(s):

Richard Boynton ◽

Homayon Aryan ◽

Walker Simon ◽

Michael Balikhin

Keyword(s):

Machine Learning ◽

Solar Wind ◽

System Identification ◽

Moving Average ◽

Training Data ◽

Machine Learning Techniques ◽

Time Lags ◽

Inner Magnetosphere ◽

Geomagnetic Indices ◽

The Waves

<p>This research develops forecast models of the spatiotemporal evolution of emissions throughout the inner magnetosphere between L=2-6 and at all MLT. The system identification, or machine learning, technique based on Nonlinear AutoRegressive Moving Average eXogenous (NARMAX) models is employed to deduce the forecasting models of the lower band chorus, Hiss, and magnetosonic waves using solar wind and geomagnetic indices as inputs. It is difficult to develop machine leaning based spatiotemporal models of the waves in the inner magnetosphere as the data is sparse and machine learning techniques require large amounts of data to deduce a model. To solve this problem, the spatial co-ordinates at the time of the measurements are included as inputs to the model along with time lags of the solar wind and geomagnetic indices, while the measurement of the waves by the Van Allen Probes are used as the output to train the models. The estimates of the resultant models are compared with separate data to the training data to assess the performance of the models. The models are then used to reconstruct the spatiotemporal waves over the inner magnetosphere, as the waves respond to changes in the solar wind and geomagnetic indices. &#160;</p>

Download Full-text

Machine Learning, Probabilistic Inference, System Identification and Control

IFAC Proceedings Volumes ◽

10.3182/20120711-3-be-2027.00431 ◽

2012 ◽

Vol 45 (16) ◽

pp. 1275

Author(s):

Carl Edward Rasmussen

Keyword(s):

Machine Learning ◽

System Identification ◽

Probabilistic Inference ◽

Inference System ◽

And Control

Download Full-text

Probabilistic Approach for System Identification using Machine Learning

2019 International Conference on Information Science and Communications Technologies (ICISCT) ◽

10.1109/icisct47635.2019.9012025 ◽

2019 ◽

Author(s):

Shokhjakhon Abdufattokhov ◽

Behzod Muhiddinov

Keyword(s):

Machine Learning ◽

System Identification ◽

Probabilistic Approach

Download Full-text

Deep Learning Methods for System Identification of UAV

10.31224/osf.io/adr97 ◽

2020 ◽

Author(s):

Brilian Putra Amiruddin

Keyword(s):

Machine Learning ◽

Deep Learning ◽

System Identification ◽

Unmanned Aerial Vehicles ◽

Control Systems ◽

Recurrent Neural Networks ◽

Excellent Performance ◽

Learning Methods ◽

Life Problems ◽

Novel Algorithms

Nowadays, deep learning is the most prominent subjectin the machine learning field. With the bloom of researchers in this field, numerous novel algorithms are used to solve everyday life problems. The control systems field is one of the subjects that get many impacts of machine learning emergence. System identification of Unmanned Aerial Vehicles (UAV) is one of the control systems problems that could be solved by using deep learning methods. In this paper, Recurrent Neural Networks (RNNs) are applied toidentify the system of UAV. Three different models of Deep RNNs have been tried, and the results implied that the RNNs-1 was giving more excellent performance both on the testing MSE and RMSE with the values equal to 0.0006 and 0.0242, successively.

Download Full-text

Modelling COVID-19 Pandemic Dynamics Using Transparent, Interpretable, Parsimonious and Simulatable (TIPS) Machine Learning Models: A Case Study from Systems Thinking and System Identification Perspectives

10.1101/2021.11.01.21265653 ◽

2021 ◽

Author(s):

Hua-Liang Wei ◽

Stephen A Billings

Keyword(s):

Machine Learning ◽

System Identification ◽

Systems Thinking ◽

Systems Engineering ◽

Moving Average ◽

A Priori ◽

Learning Models ◽

New Findings ◽

Machine Learning Models

Since the outbreak of COVID-19, an astronomical number of publications on the pandemic dynamics appeared in the literature, of which many use the susceptible infected removed (SIR) and susceptible exposed infected removed (SEIR) models, or their variants, to simulate and study the spread of the coronavirus. SIR and SEIR are continuous-time models which are a class of initial value problems (IVPs) of ordinary differential equations (ODEs). Discrete-time models such as regression and machine learning have also been applied to analyze COVID-19 pandemic data (e.g. predicting infection cases), but most of these methods use simplified models involving a small number of input variables pre-selected based on a priori knowledge, or use very complicated models (e.g. deep learning), purely focusing on certain prediction purposes and paying little attention to the model interpretability. There have been relatively fewer studies focusing on the investigations of the inherent time-lagged or time-delayed relationships e.g. between the reproduction number (R number), infection cases, and deaths, analyzing the pandemic spread from a systems thinking and dynamic perspective. The present study, for the first time, proposes using systems engineering and system identification approach to build transparent, interpretable, parsimonious and simulatable (TIPS) dynamic machine learning models, establishing links between the R number, the infection cases and deaths caused by COVID-19. The TIPS models are developed based on the well-known NARMAX (Nonlinear AutoRegressive Moving Average with eXogenous inputs) model, which can help better understand the COVID-19 pandemic dynamics. A case study on the UK COVID-19 data is carried out, and new findings are detailed. The proposed method and the associated new findings are useful for better understanding the spread dynamics of the COVID-19 pandemic.

Download Full-text