Illustrations

Several subjects

You can find several kinds of pictures on this page:

Machine Learning and Control

The objective is the data driven control of a (unknown) system, by using statistical (unsupervised machine) learning.

Lorenz' system

In black, the uncontrolled trajectory of the Lorenz' system, during the learning stage. In red, the controlled response. Only the X3 component is observed, while the control is done on the X1 component.

Lorenz' system

Same, but only the X1-component is drawn.


Non-Uniform Dynamic Mode Decomposition

If you have a dataset which is not time resolved (in the sense of Nyquist) or too big, you can still inexpensively extract spectral informations

3D DMD Mode

Comparison of slices of the DMD and the NUDMD modes. Only 0.0001% of the original information was necessary for the NUDMD to give results!


Sample of 3D Computations

These pictures are juste a few samples !

Because 3D DNS dataset are enlightful for understanding experimental 2D dataset, here is some 3D DMD mode

3D DMD Mode

and 3D LCS of a cavity flow.
LCS in a cavity flow

Exploration of datasets

Here is some pictures of tools for exploring dataset in fluid mechanics. You can directly manipulate the dataset : Direct manipulation of a dataset
Or you can interact with the dataset : Cutting plane oriented with a tangible stylus

If you are interested in CHI for the exploration of datasets, you can also watch these videos !






Lorenz

Dr. Florimond Guéniat
Contact : contact[at]gueniat.fr
Tel : +1/(217) 721-0404
Address : 319 XPACC, Coordinated Science Laboratory, Univ. Illinois at Urbana-Champaign, 1308 W. Main Street, Urbana, IL 61801-2307 (US)

Univ. Illinois at Urbana Champaign