AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Thermostat-assisted continuously-tempered Hamiltonian Monte Carlo for Bayesian learning

Rui Luo, Jianhong Wang, Yaodong Yang, Jun WANG, Zhanxing Zhu

NeurIPS 2018 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Hamiltonian Monte Carlo · 6 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use TACT-HMC instead of the strongest reported baselineMLP classification · EMNIST Balanced, 0%, 20%, and 30% permuted labelsaccuracy: 84.53% -> 84.85%; 82.64% -> 82.95%; 81.70% -> 81.77%Improved
Use TACT-HMC instead of the strongest reported baselineRNN classification · Fashion-MNIST, 0%, 20%, and 30% permuted labelsaccuracy: 90.25% -> 90.84%; 88.91% -> 89.61%; 88.58% -> 89.01%Improved
Use TACT-HMC instead of the strongest reported baselineCNN classification · CIFAR-10, 0%, 20%, and 30% permuted labelsaccuracy: 76.60% -> 78.93%; 73.87% -> 74.88%; 71.79% -> 73.22%Improved
Remove continuous tempering while retaining thermostats (ablation)TACT-HMC · Synthetic multimodal distributions with noisy gradientssampling coverageHurt
Remove thermostatting while retaining tempering (ablation)TACT-HMC · Synthetic multimodal distributions with noisy gradientssampling accuracyHurt
Use TACT-HMC rather than SGHMCTACT-HMC · 1D Gaussian mixtureeffective sample size (ESS): 2.1096×10^4; roughly 60.2% of the value for SGHMCHurt
Use TACT-HMC rather than SGNHTTACT-HMC · 1D Gaussian mixtureeffective sample size (ESS): 2.1096×10^4; 50.9% of that for SGNHTHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy