5 ms·

Translated into a human language, this means what exactly?

steve76 5y ago
lumost 5y ago
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 5y ago
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 5y ago
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 https://www.youtube.com/watch?v=J3xLuZNKhlY
- evanb 5y ago
  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 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 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.

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 5y ago

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 5y ago

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 https://www.youtube.com/watch?v=J3xLuZNKhlY

evanb 5y ago
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 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 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.