Speaker
Daniel Lukic
(Graz University of Technology)
Description
The main goal of the project is to find a machine learning approximation for the kinetic energy functional of orbital-free density functional theory,
\begin{equation}
T[n] = \int \tau[n] \,\mathrm{d}x,
\end{equation}
where the function $\tau[n]$ is represented using a feed forward neural network. Since it is known that the function $\tau$ is translationally invariant and non-local, i.e. a function of the values of $n$ at various positions $x$, the structure of a convolutional neural network seems like a reasonable choice.
Primary authors
Prof.
Andreas Hauser
(Graz University of Technology - Institute of Experimental Physics)
Daniel Lukic
(Graz University of Technology)
Dr
Meyer Ralf
