JANUS: A bit-wise reversible integrator for N-body dynamics · HackerPapers discussions

JANUS: A bit-wise reversible integrator for N-body dynamics — mentioned by imurray

Mon, 02 Jun 2025 09:53:35 GMT

I'm sceptical about the energy motivation, but there are multiple reasons why making invertible deep learning architectures can be interesting or useful. Cf, a series of workshops from 2019-2021: https://invertibleworkshop.github.io/ Since then diffusion models have been popular. Generating from these can be seen as a special case of a continuous time normalizing flow, and so (in theory) is a reversible computation. Although the distilled/fast generation that's run in production is probably not that! Simulating differential equations is not usually actually reversible in practice due to…

Paper and discussion history. A mention is not an endorsement.

JANUS: A bit-wise reversible integrator for N-body dynamics — mentioned by antognini

Sun, 18 Feb 2018 23:16:16 GMT

Hanno Rein and Daniel Tamayo are pretty legit dynamicists. For anyone interested in doing calculations like these themselves, they've been working on this amazing N-body library called Rebound: https://github.com/hannorein/rebound One of my all-time favorite figures is Fig. 1 of this paper of theirs describing one of the integrators used in Rebound: https://arxiv.org/abs/1704.07715

Paper and discussion history. A mention is not an endorsement.

JANUS: A bit-wise reversible integrator for N-body dynamics — mentioned by antognini

Sat, 06 May 2017 22:41:02 GMT

Hanno Rein, one of the experts in the field, just published a new algorithm which is exactly time-reversible for floating point numbers. Figure 1 of the paper is pretty incredible: https://arxiv.org/abs/1704.07715

Paper and discussion history. A mention is not an endorsement.