A new arXiv preprint sets out to test whether Lagrangian and Hamiltonian neural networks can be applied to a dissipative system. The authors focus on a case where the Lagrangian, Hamiltonian, and total energy all have explicit time dependence, a condition that goes beyond many standard physics-informed modeling benchmarks.
The abstract describes the investigation but stops short of stating findings. It notes that the researchers "consider these neural" models, with the description cut off in the available text. As a result, the preprint's contribution is currently framed as a question of applicability rather than a demonstrated outcome.
For readers tracking physics-informed machine learning, the significance is in the choice of system: dissipation plus explicit time dependence is a known stress test for models built around conserved quantities. Whether the networks succeed, fail, or require modification is not clear from the abstract alone.