ML-physics interface: neural-net wave-functions (Carleo); learning Hamiltonians; symbolic-regression; equivariant networks (Cohen-Welling); ML-accelerated lattice-QCD; differentiable physics; Hamiltonian-NN; data-driven turbulence.
machine-learning-physics
Carleo neural-net wave functions
Carleo-Troyer 2017 'Solving the Quantum Many-Body Problem with Artificial Neural Networks'. RBM ansatz for ground-state. Generalises to…
Symbolic regression for physics laws
Schmidt-Lipson 2009 'Distilling Free-Form Natural Laws'; Cranmer 2020 PySR. Discover symbolic equations from data via genetic-programming…
Equivariant neural networks (Cohen-Welling)
Cohen-Welling 2016 / Bronstein 2021 'Geometric Deep Learning'. Symmetry-preserving NN architectures (rotation / translation / SE(3)). Used…
Hamiltonian + Lagrangian neural networks
Greydanus 2019 HNN; Lutter 2019 LNN. NN learns Hamiltonian-fn or Lagrangian-fn directly; conservation-laws built-in. Foundation for…
Differentiable physics engines
Hu 2019 DiffTaichi / Werling 2021. Physics-simulators with auto-diff backward-pass for gradient-based control + parameter-estimation +…
ML-accelerated lattice QCD
Albergo 2019 normalising-flow for lattice-QFT sampling; Gorinski 2024 phase-space-scaling. Reduce critical-slowing-down. Bridges ML +…
Physics-informed neural networks (PINN)
Raissi-Perdikaris-Karniadakis 2019 PINN. NN trained to satisfy PDE in residual + boundary-conditions in loss. Solves forward + inverse PDE…
Quantum machine learning (parameterised quantum circuits)
Variational-quantum-eigensolver (Peruzzo 2014) + QAOA (Farhi 2014) + quantum-NN. PQC train via classical-loop. NISQ-era applications.…
Turbulence ML models
Maulik-San 2018 ML closures for LES; Beck 2019 RANS-corrections. Data-driven closure models for unresolved-scales. Bridges fluid-dynamics +…
Auto-discovery of conservation laws
Liu-Tegmark 2021 'Machine-learning conservation laws from differential equations'. NN-based identification of conserved-quantities from…
Variational autoencoder for physics
Iten-Renner 2020 'Discovering physical concepts with neural networks'. VAE-latent-space recovers Copernican-helio-centric / quantum-state…
Graph neural networks for molecular dynamics
Schütt 2017 SchNet / Batzner 2022 NequIP / Gilmer 2017 MPNN. GNN for molecular-energy + force prediction at DFT-quality. Foundation of…
NN potentials (Behler-Parrinello 2007)
J Behler-M Parrinello 2007 NN-PES; ANI (Smith 2017) + SchNet (Schutt 2018); modern NequIP equivariant; chemical-accuracy reactive-MD.
AlphaFold 2 (Jumper 2021)
Jumper-DeepMind 2021 attention-based protein structure prediction; ~1A median accuracy; modern AF3 multi-modal 2024 + biology revolution.
Neural ODE (Chen 2018)
Chen-Rubanova-Bettencourt-Duvenaud 2018 'Neural ODEs'; continuous-depth NN; modern Lagrangian + Hamiltonian neural-networks.
PINN (Raissi 2019)
Raissi-Perdikaris-Karniadakis 2019 physics-informed-neural-networks; PDE-constrained loss; modern weak-form variants + multiphysics.
ML jet tagging (CMS 2014)
CMS 2014 + ATLAS 2017 deep-learning b-tagging at LHC; modern transformer + ParticleNet; replaced cut-based discriminators with 2-3x…
SINDy (Brunton 2016)
S Brunton-J Proctor-N Kutz 2016 sparse-identification-of-nonlinear-dynamics; modern PDE-FIND + symbolic-regression Schmidt-Lipson 2009.