no code implementations • 15 Dec 2023 • Carlos E. Pérez De Jesús, Alec J. Linot, Michael D. Graham
In this work we exploit the symmetries of the Navier-Stokes equations (NSE) and use simulation data to find the manifold where the long-time dynamics live, which has many fewer degrees of freedom than the full state representation, and the evolution equation for the dynamics on that manifold.
no code implementations • 10 Oct 2023 • C. Ricardo Constante-Amores, Alec J. Linot, Michael D. Graham
Additionally, we explore a modified approach where the system alternates between spaces of states and observables at each time step -- this approach no longer satisfies the linearity of the true Koopman operator representation.
1 code implementation • 1 May 2023 • Kevin Zeng, Carlos E. Pérez De Jesús, Andrew J. Fox, Michael D. Graham
Analysis of gradient descent dynamics for this architecture in the linear case reveals the role of the internal linear layers in leading to faster decay of a "collective weight variable" incorporating all layers, and the role of weight decay in breaking degeneracies and thus driving convergence along directions in which no decay would occur in its absence.
no code implementations • 28 Jan 2023 • Alec J. Linot, Kevin Zeng, Michael D. Graham
The high dimensionality and complex dynamics of turbulent flows remain an obstacle to the discovery and implementation of control strategies.
no code implementations • 11 Jan 2023 • Alec J. Linot, Michael D. Graham
For comparison, we show that the models outperform POD-Galerkin models with $\sim$2000 degrees of freedom.
no code implementations • 20 Nov 2022 • Charles D. Young, Michael D. Graham
A common problem in time series analysis is to predict dynamics with only scalar or partial observations of the underlying dynamical system.
no code implementations • 29 Oct 2022 • Carlos E. Pérez De Jesús, Michael D. Graham
At a model dimension of five for the pattern dynamics, as opposed to the full state dimension of 1024 (i. e. a 32x32 grid), accurate predictions are found for individual trajectories out to about two Lyapunov times, as well as for long-time statistics.
no code implementations • 1 May 2022 • Kevin Zeng, Alec J. Linot, Michael D. Graham
We show that the ROM-based control strategy translates well to the true KSE and highlight that the RL agent discovers and stabilizes an underlying forced equilibrium solution of the KSE system.
no code implementations • 29 Mar 2022 • Alec J. Linot, Joshua W. Burby, Qi Tang, Prasanna Balaprakash, Michael D. Graham, Romit Maulik
We present a data-driven modeling method that accurately captures shocks and chaotic dynamics by proposing a novel architecture, stabilized neural ordinary differential equation (ODE).
no code implementations • 31 Aug 2021 • Alec J. Linot, Michael D. Graham
Then the ODE, in these coordinates, is approximated by a NN using the neural ODE framework.
1 code implementation • 12 Aug 2021 • Daniel Floryan, Michael D. Graham
We demonstrate this approach on several high-dimensional systems with low-dimensional behavior.
no code implementations • 9 Apr 2021 • Kevin Zeng, Michael D. Graham
Many systems of flow control interest possess symmetries that, when neglected, can significantly inhibit the learning and performance of a naive deep RL approach.
1 code implementation • 20 Dec 2019 • Alec J. Linot, Michael D. Graham
A data-driven framework is developed to represent chaotic dynamics on an inertial manifold (IM), and applied to solutions of the Kuramoto-Sivashinsky equation.