[HN Gopher] Lattice Gauge Equivariant Convolutional Neural Networks
___________________________________________________________________
Lattice Gauge Equivariant Convolutional Neural Networks
Author : amai
Score : 25 points
Date : 2022-02-03 15:44 UTC (2 days ago)
(HTM) web link (journals.aps.org)
(TXT) w3m dump (journals.aps.org)
| Llamamoe wrote:
| Translated into a human language, this means what exactly?
| lumost wrote:
| They defined a neural network operation which can in principal
| learn to approximate certain equations used in Nuclear physics
| and other quantum field theories.
|
| Approximating QFT/Nuclear physics is a very difficult task
| given current algorithms, so this opens potential doors for
| learning "faster" approximations which allow simulating more
| complex systems. Of course, defining the error in this
| approximation becomes quite difficult. They claim the following
| in their paper.
|
| 1. A theoretical proof that the new operation can approximate
| the relevant equations.
|
| 2. An evaluation of the model trained against a traditional
| simulation on both equal sized and "larger sized" problems.
| Unfortunately they only summarize this as a "High degree of
| accuracy" and present charts. I'd love to see a summary table
| of results with a final numeric accuracy evaluation.
| savant_penguin wrote:
| They summarize it in the abstract
|
| "At the heart of this network structure is a novel
| convolutional layer that preserves gauge equivariance while
| forming arbitrarily shaped Wilson loops in successive bilinear
| layers."
|
| Much simpler
| tempay wrote:
| I've found a surprising lack of accessible content about
| lattice QCD however it's about simulating the smallest scales
| in the universe. The calculations are computationally
| infeasible however you can get useful results from
| approximiting them in a grid of cells and seeing how it evolves
| over time. This is still incredibly complex and typically needs
| the largest supercomputers they can find. Personally I'm a
| little sceptical that it will be possible to use neural network
| based approximations to produce physically meaningful results
| with well understood errors but time will tell.
|
| If you're so inclined, I think this video gives a nice
| description of lattice QCD:
|
| https://www.youtube.com/watch?v=J3xLuZNKhlY
| evanb wrote:
| When my collaborators and I published our LQCD computation of
| the nucleon axial coupling I wrote this
| http://evanberkowitz.com/2018/05/30/gA.html though I did not
| get into the technical details of gauge invariance.
|
| We can use neural networks in some steps but not in others.
| The main use seems to be in the generation of gauge
| configurations (states of the Markov chains we use). We can
| use neural networks there because we alway do our Metropolis-
| Hastings accept/reject step according to the 'true' action,
| which corrects any numerical errors introduced by the ML
| approximation.
|
| There are different approaches for this. I've been working on
| incorporating ML into HMC; see eg.
| https://arxiv.org/abs/2006.11221. For an approach based on
| normalizing flows, search for Shanahan eg al.
|
| Happy to answer questions, but it's late and I'm about to go
| to sleep. Feel free to ask away and I will do my best in the
| morning. Or email me.
___________________________________________________________________
(page generated 2022-02-05 23:01 UTC)