Alexander Falk
@alfa98.bsky.social
PhD Candidate/University Assistant @ Graz University of Technology
If you’re using gradient-based MCMC sampling, this might be for you 🔥
We extend ULA with momentum (inspired by Polyak’s heavy ball algorithm). The result: faster sampling ⚡ with provable ergodicity and bias bounds in Wasserstein-2.
👉 Go to: arxiv.org/abs/2510.06723
#MCMC #Langevin #Sampling #ML
We extend ULA with momentum (inspired by Polyak’s heavy ball algorithm). The result: faster sampling ⚡ with provable ergodicity and bias bounds in Wasserstein-2.
👉 Go to: arxiv.org/abs/2510.06723
#MCMC #Langevin #Sampling #ML
October 9, 2025 at 7:39 AM
If you’re using gradient-based MCMC sampling, this might be for you 🔥
We extend ULA with momentum (inspired by Polyak’s heavy ball algorithm). The result: faster sampling ⚡ with provable ergodicity and bias bounds in Wasserstein-2.
👉 Go to: arxiv.org/abs/2510.06723
#MCMC #Langevin #Sampling #ML
We extend ULA with momentum (inspired by Polyak’s heavy ball algorithm). The result: faster sampling ⚡ with provable ergodicity and bias bounds in Wasserstein-2.
👉 Go to: arxiv.org/abs/2510.06723
#MCMC #Langevin #Sampling #ML