HackerPapers

arXiv:1906.01563cs.NE

Hamiltonian Neural Networks

Sam Greydanus, Misko Dzamba, Jason Yosinski

Abstract

Even though neural networks enjoy widespread use, they still struggle to learn the basic laws of physics. How might we endow them with better inductive biases? In this paper, we draw inspiration from Hamiltonian mechanics to train models that learn and respect exact conservation laws in an unsupervised manner. We evaluate our models on problems where conservation of energy is important, including the two-body problem and pixel observations of a pendulum. Our model trains faster and generalizes better than a regular neural network. An interesting side effect is that our model is perfectly reversible in time.

arXiv comments
Conference paper at NeurIPS 2019. Main paper has 8 pages and 5 figures
Journal reference
DOI
Subjects
cs.NENeural and Evolutionary Computing

On Hacker News

Linked by 3 identifiable commenters in 3 comments across 3 threads, from Feb 2020 to Jul 2020.

Comments linking this paper, per year.

Discussed in comments3 threads

  1. AI in physics: are we facing a scientific revolution? 1 comment

    • There were some interesting talks on neural differentiation applied to physics at ICLR. You probably saw: https://arxiv.org/abs/1906.01563 Very fun!

      ylemView on HNParent context

  2. Teaching physics to neural networks removes 'chaos blindness' 1 comment

    • Hamiltonian NNs are not a new thing. There was a NIPS 2019 paper [0] that attempted to do that same for some toy problems. In general the idea of including model or context-based information into neural networks goes along the line of Kahneman's System I and System II of the…

      seesawtronView on HNParent context

  3. Sometimes, requiring explanability is counterproductive 1 comment

    • It's actually extraordinarily difficult to get a DNN model to accurately predict planetary motion over arbitrary time horizons. RNNs have a tendency to either decay or blow-up. The only way I know to make neural nets work well for this problem is to build in lots of physics into…

      shoyerView on HNParent context

Submitted to Hacker News

Connected through shared HN threads, not inferred similarity or a recommendation.