AI Paper Atlas

Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes

Random measure priors in Bayesian recovery from sketches

Mario Beraha, Stefano Favaro, Matteo Sesia

JMLR 2024 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Nonparametric and PAC-Bayes Bayesian learning

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 PYP prior instead of DP prior for frequency recovery on Zipf dataBNP frequency recovery from sketches · Zipf synthetic data (n=500,000)mean absolute errorImproved
estimate prior parameters by minimizing recovery error on first 10,000 samples instead of the likelihood-free approachPYP frequency recovery · Zipf synthetic data (n=500,000)mean absolute errorImproved
use PYP prior instead of DP prior on the Gutenberg bigramsBNP frequency recovery · Gutenberg 2-grams (1,700,000)mean absolute errorImproved
increase hash width J under the PYP estimator at c=1.82 (ablation)PYP frequency estimator · Zipf data c=1.82mean absolute errorHurt
apply DP prior to heavy-tailed Zipf data rather than DP-generated data (ablation)DP frequency recovery · Zipf synthetic data vs DP synthetic datamean absolute errorHurt
use DP-prior cardinality estimator on data with power-law tails and narrow hash width (ablation)BNP cardinality recovery · Gutenberg, SARS-CoV-2 16-mers, IP addressescardinality estimation accuracyMixed

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