Moritz Reh
34 posts
Moritz Reh
@RehMoritz
PhD student in Machine Learning & Quantum Many-Body Physics @GaerttnerGroup @UniHeidelberg
Heidelberg, Germany Katılım Mart 2021
169 Takip Edilen113 Takipçiler
It will definitely take some time to realise.. thanks for everyone being part of my Ph.D. journey! Looking forward for what comes next 🚀
Correlated Quantum Materials@CQM_Aalto
Congratulations to Rouven Koch (@koch_rouven), Ph.D. researcher @CQM_Aalto, for a wonderful Ph.D. defense. It was great having Prof. Daniele Passerone acting as the opponent.
English
For a longer, more thorough analysis: arxiv.org/abs/2404.12323
And for a rigorous theoretical perspective: arxiv.org/abs/2404.12320
English
Happy to report on a new work that is (finally) out:
We demonstrate that area laws are not restricted to quantum entropies but can also be observed in differential entropies of phase-space distributions!
We showcase this exemplarily for a spin-1 BEC: arxiv.org/abs/2404.12321
GIF
English
What is advantageous about phase-space entropies is that estimating their value is feasible for a moderate number of samples, in contrast to full-state tomography so implementing the proposal experimentally seems feasible, even if we do not assume a specific functional form.
English
The main hurdle that we needed to overcome was realizing that to first order, differential entropies are extensive - the area law is only visible as a deviation from this extensive scaling.
English
While we observe a logarithmic scaling for early times, we find a transition to an extensive (volume-law) scaling for late times, reminiscent of thermalization, just as one would expect for the entanglement entropy.
English
Moritz Reh retweetledi
I have been appointed an associate professor of computational physics at @ETH_physics. I'll have PhD and postdoc positions, so get in touch if you are interested in quantum many-body physics, ML, condensed matter, or quantum info/computing
English
Moritz Reh retweetledi
We have used supervised learning to estimate entanglement entropy and quantum mutual information in spin chains using very few experimental samples. On top, the method also estimates the statistical error. Comments welcome! arxiv.org/abs/2309.07556
English
Moritz Reh retweetledi
Thanks to everyone who helped make Machine Learning for Quantum Many-Body Systems a success @Perimeter ! We can't virtually recreate the coffee & blackboard chats, but the recorded talks are all online on PIRSA:
pirsa.org/C23002
English
Moritz Reh retweetledi
'Variational #MonteCarlo approach to partial differential equations with #neuralnetworks' by @RehMoritz @magaerttner @GaerttnerGroup @structures_hd hits 500 downloads! bit.ly/3OSyVNp #machinelearning #PDEs #statphys #compchem #HPC #AI #fluids #dynamicalsystems #quantum
English
Which network architectures capture the properties of quantum systems optimally? We systematically compare different NQS ansatz functions on benchmark problems and identify the details of symmetrization as a key to better performance:
arxiv.org/abs/2301.06788
English
Happy to see our work published in MLST!
Machine Learning: Science and Technology@MLSTjournal
Great new work by @RehMoritz @magaerttner @GaerttnerGroup @structures_hd - 'Variational #MonteCarlo approach to partial differential equations with #neuralnetworks' - bit.ly/3OSyVNp #machinelearning #PDEs #statphys #compchem #HPC #AI #fluids #dynamicalsystems #quantum
English
Moritz Reh retweetledi
How can machine learning be used for an efficient state characterization for quantum simulators? Read more about recent research breakthroughs in the short news abstract by @SchmaleTobias, @RehMoritz & @magaerttner (@GaerttnerGroup) on their latest works:
#2022-10-20" target="_blank" rel="nofollow noopener">structures.uni-heidelberg.de/news.php?twt_c…
English
Paper: arxiv.org/abs/2206.01927
Code: github.com/RehMoritz/vmc_…
4/4
English
However, outside of NQS this approach was yet unexplored. We show that the theory and derivations for the PDE case correspond 1-to-1 to those of NQS and apply our normalizing-flow based method to high-dimensional Fokker-Planck equations.
3/4
English
Solving PDEs with neural nets is becoming increasingly popular, but hinges on solving non-convex optimizations. Combining tools from neural quantum states with NFs we avoid iterative optimization & the curse of dimensionality of discretization schemes: arxiv.org/abs/2206.01927 🧵
English
Moritz Reh retweetledi
[beware, Tweet contains book placement]
What happens if you combine the best experts in the field with hard-working and ambitious students? You get a Book. We’re excited to present “Modern applications of machine learning in quantum sciences”!
arxiv.org/abs/2204.04198
English
🥳🎉
Quantum Information & Quantum Simulations @Jena@GaerttnerGroup
Our work on dissipative quantum dynamics using neural nets is now published in @PhysRevLett! We derive an analytic expression for the Lindblad dynamics on the network's variational manifold, which can be efficiently estimated. Find it here: journals.aps.org/prl/abstract/1…
ART