AI Paper Atlas

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

Nearest Neighbor Dirichlet Mixtures

Shounak Chattopadhyay, Antik Chakraborty, David B. Dunson

JMLR 2023 · 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 NN-DM instead of DP-MC for credible intervalsNN-DM · Bivariate densities (n=500)frequentist coverage: MG 0.92 vs 0.53; MVC 0.69 vs 0.47; SN 0.92 vs 0.39Improved
use NN-DM instead of DP-VB for credible intervalsNN-DM · Bivariate densities (n=500)frequentist coverage: MVC 0.69 vs 0.18; MG 0.92 vs 0.56Improved
use NN-DM instead of DP-MC/DP-VB with similar interval widthNN-DM · Univariate claw and sawtoothfrequentist coverage: CW 0.89 vs 0.06 and 0.18; ST 0.91 vs 0.64 and 0.07Improved
use NN-DM instead of DP-MC with default hyperparameters on sawtooth densityNN-DM · Sawtooth density n=500density estimateImproved
use NN-DM instead of DP-VB at p=50NN-DM · High-dim p=50, n=1000out-of-sample log-likelihood: MVC -1.31 vs -1.34; T -1.84 vs -1.86; MG -1.75 vs -1.75No clear effect
use NN-DM with default hyperparameters instead of DP-MCNN-DM · Runtime p=1 and p=4, n=1500runtime (s): 3.3 vs 99.4 (p=1); 16.4 vs 1618.1 (p=4)Improved
use NNDM package instead of dirichletprocess packageNN-DM · n=1500, p=4runtime (s): 3.52 vs 1068.48Improved
vary number of neighbors k from 2 to 50 (ablation)NN-DM · Univariate SP and bivariate MG, n=200out-of-sample log-likelihood: best k=9 (SP) and k=12 (MG)No clear effect
cross-validated Psi_0 instead of default hyperparameters (ablation)NN-DM · High-dim p=50out-of-sample log-likelihoodNo clear effect
use NN-DM instead of KDE as dimension grows from 2 to 6NN-DM · Multivariate densitiesL1 error: KDE error rises by factors of about 5 (n=200) and 7 (n=1000)Improved
use NN-DM instead of DP-VB for pulsar classificationNN-DM · HTRU2 pulsar classificationsensitivity and specificityImproved
use NN-DM instead of DP-MC for pulsar classification at 1800 training pointsNN-DM · HTRU2 pulsar classificationcomputation time: about 13 minutes vs approximately 5 hoursImproved

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