Optimizing training data via a differentiable SCM yields climate emulators that outperform those trained on six standard ScenarioMIP pathways while using less data and isolating distinct forcing responses.
hub
Christopher Bishop.Pattern Recognition and Machine Learning
21 Pith papers cite this work. Polarity classification is still indexing.
21
Pith papers citing it
hub tools
citation-role summary
background 3
citation-polarity summary
roles
background 3polarities
background 3representative citing papers
Targeted perturbations in the Aurora AI model can steer Hurricane Sandy's trajectory by more than 500 km after seven days via amplification in sensitive regions identified by FTLE and wave activity diagnostics.
Deep-Koopman-KANDy recovers symbolic Koopman dictionaries post-training by replacing the encoder and decoder with KANs and applying a level-set construction with chain-rule gradients, achieving high recall on Lorenz and expected behavior on other maps.
Cast3 translates NWP principles into a data-driven model using cubed-sphere grids, super-ensembles, and generative nudging to achieve state-of-the-art ensemble predictions that outperform baselines.
A new linearized dynamics based on the anti-symmetric part of the stability matrix preserves phase space volume for non-Hamiltonian chaotic systems within a classical density matrix framework.
Sparse hierarchies of K coupled gyrostats (M=2K+1 modes, no linear feedback) possess exactly (M+1)/2 quadratic invariants recoverable as Casimirs of a constructible non-canonical Poisson structure.
FPPF uses a learned conditional generative proposal approximating the optimal proposal in particle filters, with tractable likelihoods for Bayesian updates and localization for high dimensions, outperforming baselines on nonlinear non-Gaussian systems.
A kernel-based data-driven optimization method computes optimal perturbations to control the spectrum of transfer operators in high-dimensional dynamical systems.
PIDM-DP integrates Dormand-Prince ODE solving into DDPM denoising with scheduled physics guidance to reconstruct chaotic states, reporting up to 15.4x RMSE gains over baselines on five systems including stiff cases.
New dynamical series from Morse flow loops in fibered knot complements conjectured to equal BPS q-series encoding colored Jones polynomials, with proof for braid-homogeneous knots.
SURGE is an unbiased particle filter that fuses diffusion-model simulations with noisy observations via sequential Monte Carlo reweighting over diffusion trajectories.
A framework recasts multivariate time series forecasting as scalar regression problems that tabular prior-fitted networks can solve zero-shot while addressing inter-channel interactions.
MASF redesigns the forward diffusion process to align with measurements, yielding a theoretically grounded likelihood score and up to 28.2x speedup on O(10^5)-dimensional Kolmogorov flow under sparse and nonlinear observation operators.
HealDA supplies ML-based initial conditions for AI weather models that produce forecasts trailing ERA5-initialized runs by less than one day of effective lead time, with the skill gap arising mainly from initial error size.
In the Lorenz '96 system, stochastic parameterizations with temporal persistence improve early ensemble spread growth and spread-error consistency without increasing long-term variance.
GIFT fine-tunes deep RL policies with a stability-focused reward to improve global stability while preserving task performance.
A quantum echo-state network is implemented on NISQ superconducting qubits and shown to predict long chaotic trajectories from the Lorenz system with memory persisting over 100 times the median T1/T2 time.
WSINDYc-MPC identifies governing dynamics more robustly than benchmarks under high noise, enabling longer prediction horizons and lower tracking errors in fusion, drone, chaos, and aircraft control tasks.
Neural network learning opacity stems from three dynamical complexity properties in training, rendering some sources of opacity irreducible.
Divide-and-conquer modeling using scenario-specific techniques reaches a public score of 79.63 on the CTF-4-Science Lorenz benchmark.
On the Lorenz model, 4DVAR tracks the true trajectory better than the Ensemble Kalman Filter for 20% initial errors but both fail for 40% errors with only three observations; performance improves when all variables are observed.
citing papers explorer
No citing papers match the current filters.