Innovation
Methods, ideas, or system contributions that make the work stand out.
Gaussian cloud sampler
Picard HMC
first-order rejection sampling
likelihood correction gadgets
log-Sobolev inequality
High-accuracy simulation of Picard HMC, part I: Gaussian cloud correction
🤖 AI Summary
The original TLDR content provided is incomplete, containing only fragmented phrases such as "this project adopts/uses," which precludes the extraction of core information. The following serves as a generalized academic TLDR template for populating with specific details. To address the limitations of existing methods in a given domain, this work proposes a novel approach grounded in a specified core technology or theory. By introducing key innovative modules, the proposed method achieves its objectives through a distinct technical mechanism. Experimental evaluations demonstrate that this approach outperforms state-of-the-art baselines on standard benchmarks by measurable margins, effectively resolving the aforementioned challenges. The primary contribution lies in pioneering the application of a specific innovation to a targeted scenario, thereby establishing a new paradigm characterized by both efficiency and robustness for the broader research community.
📝 Abstract
We study the problem of sampling from a continuous density $π\propto \exp(-V)$ on $\mathbb R^d$, where $V\in C^2(\mathbb R^d)$ has a $β$-Lipschitz gradient and $π$ satisfies a logarithmic Sobolev inequality with constant $α^{-1}$, and write $κ= β/α$. We introduce the Gaussian cloud sampler, which achieves total variation accuracy $\varepsilon$ using $\widetilde O(κd^{1/5}\,\text{polylog}(1/\varepsilon))$ gradient queries in expectation. The algorithm uses first-order rejection sampling (FORS) to correct the law of smoothed Picard HMC trajectories. To do so, we represent the iterates of the ideal Picard iteration via Gaussian clouds, whose centers are never evaluated, and we develop a suite of likelihood correction gadgets which only use samples from this indirect cloud representation.
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
No related jobs found.
