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.

168,786 papers · 148 categories

Trend · papers per month

count % of corpus

8 results for “ULD”

Dimension-Independent Convergence of Underdamped Langevin Monte Carlo in KL Divergencecs.LG

New method proves dimension-free convergence for ULD in KL divergence.

problem Polynomial scaling of existing convergence guarantees in high dimensions.

method Refined KL local error framework, focusing on tr(H) instead of d.

result First dimension-free KL divergence bounds for discretized ULD.

Underdamped Langevin KL divergence Monte Carlo Gibbs sampling Hessian

2026-03-02abs ↗pdf ↗

Reverse Transition Kernel: A Flexible Framework to Accelerate Diffusion Inferencestat.ML

A new framework RTK accelerates diffusion inference by breaking down the process into fewer, more efficient subproblems.

problem Efficiently generating data from trained diffusion models using discretized reverse SDEs or ODEs.

method Developed a general RTK framework that decomposes the diffusion process into fewer, more balanced subproblems, using MALA and ULD for sampling.

result The RTK-MALA and RTK-ULD algorithms achieve faster convergence rates and lower error compared to existing methods.

Diffusion Models Inference Algorithms Reverse Transition Kernels Metropolis-Adjusted Langevin Algorithm Underdamped Langevin Dynamics

2024-05-26abs ↗pdf ↗

The Randomized Midpoint Method for Log-Concave Samplingcs.LG

New algorithm speeds up sampling from log-concave distributions.

problem Sampling from log-concave distributions efficiently.

method Markov chain Monte Carlo (MCMC) based on underdamped Langevin diffusion (ULD).

result Significantly faster than previous methods, achieving ε·D error in O(κ^7/6/ε^1/3 + κ/ε^2/3) steps.

log-concave sampling Markov chain Monte Carlo ULD Langevin diffusion Wasserstein distance

2019-09-12abs ↗pdf ↗

Breaking Reversibility Accelerates Langevin Dynamics for Global Non-Convex Optimizationmath.OC

Langevin dynamics (LD) has been proven to be a powerful technique for optimizing a non-convex objective as an efficient algorithm to find local minima while eventually visiting a global minimum on longer time-scales. LD is based on the first-order Langevin diffusion which is reversible in time. We study two variants th…

2018-12-19abs ↗pdf ↗

Underdamped Langevin MCMC with third order convergencestat.ML

New Langevin method achieves third order convergence for strongly log-concave distributions.

problem Sampling from complex distributions efficiently.

method Underdamped Langevin diffusion with third order convergence.

result Achieves 2-Wasserstein error of ε in O(√d/ε^1/3) steps under additional Lipschitz condition.

Langevin MCMC Underdamped Langevin Strongly log-concave Third order convergence Wasserstein distance

2025-08-22abs ↗pdf ↗

Unifying Model-Free Efficiency and Model-Based Representations via Latent Dynamicscs.LG

Unified Latent Dynamics unifies model-free and model-based reinforcement learning.

problem Combining the efficiency of model-free methods with the representational strengths of model-based approaches.

method Embedding state-action pairs into a latent space where the true value function is approximately linear, using synchronized updates of encoder, value, and policy networks.

result ULD achieves cross-domain competence with minimal tuning and a fraction of the parameter footprint.

reinforcement learning latent dynamics model-free model-based temporal-difference

2026-02-13abs ↗pdf ↗

Fractional Underdamped Langevin Dynamics: Retargeting SGD with Momentum under Heavy-Tailed Gradient Noisestat.ML

New method targets deep learning optima with heavy-tailed noise.

problem Heavy-tailed gradient noise biases SGDm optima.

method Fractional Underdamped Langevin Dynamics (FULD) for heavy-tailed noise.

result FULD targets Gibbs distribution matching optima of cost function.

SGDm Langevin Dynamics Heavy-tailed Noise Optimization Deep Learning

2020-02-13abs ↗pdf ↗

An Improved Analysis of Langevin Algorithms with Prior Diffusion for Non-Log-Concave Samplingcs.LG

Improved Langevin algorithms with prior diffusion achieve dimension-independent convergence for non-log-concave distributions.

problem Understanding the dimension dependency of computational complexity in high-dimensional sampling.

method Investigation of prior diffusion technique for log-Sobolev inequality target distributions.

result Modified Langevin algorithm achieves dimension-independent KL divergence convergence.

Langevin algorithms log-Sobolev inequality sampling dimension independence KL divergence

2024-03-10abs ↗pdf ↗