Constructing Neural Network Based Models for Simulating Dynamical Systems

Legaard, Christian Møldrup; Schranz, Thomas; Schweiger, Gerald; Drgoňa, Ján; Falay, Basak; Gomes, Cláudio; Iosifidis, Alexandros; Abkar, Mahdi; Larsen, Peter Gorm

doi:10.1145/3567591

Cited by 58 publications

(21 citation statements)

References 60 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…258 Directsolution models estimate the state at a specific time without integration, while time-stepper models advance it forward in time at collocation points. 259 The flexibility of the underlying structure allows species to move from spectator to consumable status. 260 A chemical reaction neural net (CRNN) is a type of neural ODE model which incorporates fundamental physical laws, such as the law of mass action and the Arrhenius law, into its structure.…”

Section: H C C H C C C C Cos( )( )mentioning

confidence: 99%

Mixtures Recomposition by Neural Nets: A Multidisciplinary Overview

Nicolle,

Deng,

Ihme

et al. 2024

J. Chem. Inf. Model.

View full text Add to dashboard Cite

Artificial Neural Networks (ANNs) are transforming how we understand chemical mixtures, providing an expressive view of the chemical space and multiscale processes. Their hybridization with physical knowledge can bridge the gap between predictivity and understanding of the underlying processes. This overview explores recent progress in ANNs, particularly their potential in the ’recomposition’ of chemical mixtures. Graph-based representations reveal patterns among mixture components, and deep learning models excel in capturing complexity and symmetries when compared to traditional Quantitative Structure–Property Relationship models. Key components, such as Hamiltonian networks and convolution operations, play a central role in representing multiscale mixtures. The integration of ANNs with Chemical Reaction Networks and Physics-Informed Neural Networks for inverse chemical kinetic problems is also examined. The combination of sensors with ANNs shows promise in optical and biomimetic applications. A common ground is identified in the context of statistical physics, where ANN-based methods iteratively adapt their models by blending their initial states with training data. The concept of mixture recomposition unveils a reciprocal inspiration between ANNs and reactive mixtures, highlighting learning behaviors influenced by the training environment.

show abstract

Section: H C C H C C C C Cos( )( )mentioning

confidence: 99%

Mixtures Recomposition by Neural Nets: A Multidisciplinary Overview

Nicolle,

Deng,

Ihme

et al. 2024

J. Chem. Inf. Model.

View full text Add to dashboard Cite

show abstract

“…Anomalous trajectories require further analysis that cannot be provided by the deep mixture of experts network. Here, it is more natural to model the measurement process as a continuous differential equation [47,48]. To this end, each future predicted trajectory y is assumed to be approximated by the following reservoir computing network [49]…”

Section: Trajectory Intention Predictionmentioning

confidence: 99%

Uncovering Drone Intentions using Control Physics Informed Machine Learning

Perrusquía

Guo

Fraser

et al. 2023

Preprint

View full text Add to dashboard Cite

Fully autonomous aerial platforms or drones are increasingly used across diverse application areas. Uncooperative drones do not announce their identity nor file flight plans and can pose a potential risk to a variety of critical infrastructures. Understanding an uncooperative drone's intention is important to assigning risk and executing countermeasures. Drones have rapidly changing design, flexible capabilities, and diverse underpinning algorithms. This makes distinguishing malicious from naive intentions across platforms difficult. Intentions are often intangible and unobservable, and a variety of tangible intention classes are often inferred as a proxy. However, inference of drone intention classes (e.g., via inverse learning) using observational data alone is inherently unreliable due to observational and learning bias. Here we develop a novel control-physics informed machine learning (CPhy-ML) that can robustly infer across intention classes, out performing over expert-defined anomaly detection and inverse learning approaches. Our CPhy-ML complementary learning couples the representation power of deep learning with the conservation laws of aerospace control models, reducing bias and instability. The results in simulation and experimentation demonstrate that we can find common intention patterns across the inferred intention classes, ranging from trajectory to reward goals. We believe that this framework can provide deeper insight into the complex nature of intention and a firm step towards to its smooth integration in current counter drone technologies. The techniques developed here can aid the enforcement of drone incursions in a time where geo-fencing is no-longer efficiently enforceable. As transport and operational autonomy is increasingly used, enforcement algorithms can also inform the design of better criminal justice system and safer autonomous systems themselves.

show abstract

“…Other ideas include synthetic data generation by simulation of a physics-based model to amplify a machine learning model input domain [13]; formulating a hybrid loss function for the learning process and penalizing the model for deviation from physical constraints [2]; pre-training a machine learning model using physics-based simulation data and fine-tuning based on a limited number of observations [14]; customizing the topology of a machine learning model by adding intermediate physics informed components [15]; designing the structure of a machine learning model using physical laws [16]; uncertainty reduction in machine learning using physics [17]; and, integrating machine learning models with first principle physical laws for solving partial differential equations [18]. A comprehensive discussion of these approaches can be found in [1,2,19].…”

Section: Introductionmentioning

confidence: 99%

Hybrid Modeling of a Multidimensional Coupled Nonlinear System With Integration of Hamiltonian Mechanics

Abbasi

Kambali

Nataraj

2023

Preprint

View full text Add to dashboard Cite

This study concerns hybrid modeling of a multidimensional coupled nonlinear system. The underlying basis for the model is derived from Hamiltonian mechanics capitalizing on the broad utility and efficiency of energy-based reasoning in modeling high-dimensional systems. The hybrid model is essentially an artificial neural network with a computational graph that is modified from conventional neural networks in a few significant ways. The first modification includes incorporating an intermediate scalar function representing the Hamiltonian learned from data. The second modification enhances input/output channels for capturing the multidimensional dynamics of the system. The main goal of such hybrid reasoning is to improve the extrapolation capability of the model by enforcing conformance with some key aspects of the underlying physics in the form of a bias. The results demonstrate that incorporating this physics-based bias into the hybrid model empowers it to produce long-term and physically plausible predictions. The proposed modeling approach also shows high scalability for energy-based modeling of multidimensional dynamic systems in general.

show abstract

Constructing Neural Network Based Models for Simulating Dynamical Systems

Cited by 58 publications

References 60 publications

Mixtures Recomposition by Neural Nets: A Multidisciplinary Overview

Mixtures Recomposition by Neural Nets: A Multidisciplinary Overview

Uncovering Drone Intentions using Control Physics Informed Machine Learning

Hybrid Modeling of a Multidimensional Coupled Nonlinear System With Integration of Hamiltonian Mechanics

Contact Info

Product

Resources

About