On-device research index

arXiv research

A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.

168,695 papers · 148 categories

Trend · papers per month

count % of corpus

4 results for “particle-mesh”

COmoving Computer Acceleration (COCA):

N

-body simulations in an emulated frame of referenceastro-ph.IM

COCA accelerates $N$ -body simulations by correcting ML errors.

problem Computational expense and limited trustworthiness of ML emulations.

method Hybrid framework combining ML and

N

-body simulator in an emulated frame of reference.

result COCA reduces emulation errors with fewer force evaluations.

N-body simulations machine learning emulation cosmological simulations particle-mesh

2024-09-03abs ↗pdf ↗

Bridging Simulators with Conditional Optimal Transportastro-ph.CO

New emulator bridges simulators using conditional optimal transport.

problem Bridging simulators with minimal distortion.

method Flow-based approach to learn likelihood transport, COT-FM for optimal matching.

result Emulator accurately captures full correction between simulators.

simulators conditional optimal transport emulator weak lensing LPT

2025-10-28abs ↗pdf ↗

CHARM: Creating Halos with Auto-Regressive Multi-stage networksastro-ph.CO

CHARM creates mock halo catalogs from dark matter density fields using neural networks.

problem Creating accurate mock halo catalogs for cosmological studies is computationally expensive.

method CHARM uses multi-stage neural spline flow networks to learn the mapping from dark matter density fields to halo catalogs.

result Mock halo catalogs have the same statistical properties as those from high-resolution N-body simulations.

halo catalog neural networks dark matter cosmological inference simulation-based inference

2024-09-13abs ↗pdf ↗

Bayesian deep learning for cosmic volumes with modified gravityastro-ph.CO

Bayesian neural networks improve cosmic parameter estimation from modified gravity simulations.

problem Estimating cosmological parameters from large-scale structure data with modified gravity.

method Implement Bayesian neural networks (BNNs) with two cases: single BLL and FullB, trained on dark matter only particle mesh

N

-body simulations.

result BNNs yield well-calibrated uncertainty estimates and accurately predict cosmological parameters for

Ω_m

and

σ_8

Bayesian neural networks modified gravity cosmic structure machine learning cosmological parameters

2023-09-01abs ↗pdf ↗