Gaussian Mixture Model
A probabilistic model formed by a weighted mixture of multiple Gaussian distributions, used to model complex multimodal distributions. It is widely applied in clustering, density estimation and related fields.
Timeline (last 90 days)
The paper controls the convergence rate of the NPMLE by sharply characterizing the bracketing entropy of finite Gaussian mixture classes under local Hellinger distance.
The paper proves that when the true mixture measure satisfies the sub-Gaussian assumption, the unconstrained NPMLE achieves the minimax optimal rate log(n)/n under squared Hellinger distance.
In separated isotropic Gaussian mixtures, the Bayesian velocity becomes locally affine, and the label-oracle residual regression benchmark has a parametric excess risk scale of K A_v(t)/N.
On Gaussian mixture models, the paper's method can approach the true target distribution.
RAPTOR works by splitting documents into text chunks, vectorizing them, performing soft clustering using Gaussian Mixture Models, calling a language model to generate abstract summaries, and recursively constructing a summary tree.
The paper uses Gaussian mixture models and full-batch gradient descent as the theoretical analysis objects.
K-means can be viewed as the hard-limit case of the EM algorithm on Gaussian mixture models, where the EM algorithm uses soft assignments while K-means uses hard assignments
All Facts (7)
The paper controls the convergence rate of the NPMLE by sharply characterizing the bracketing entropy of finite Gaussian mixture classes under local Hellinger distance.
50%UnverifiedThe paper proves that when the true mixture measure satisfies the sub-Gaussian assumption, the unconstrained NPMLE achieves the minimax optimal rate log(n)/n under squared Hellinger distance.
50%UnverifiedIn separated isotropic Gaussian mixtures, the Bayesian velocity becomes locally affine, and the label-oracle residual regression benchmark has a parametric excess risk scale of K A_v(t)/N.
50%UnverifiedOn Gaussian mixture models, the paper's method can approach the true target distribution.
50%UnverifiedRAPTOR works by splitting documents into text chunks, vectorizing them, performing soft clustering using Gaussian Mixture Models, calling a language model to generate abstract summaries, and recursively constructing a summary tree.
50%UnverifiedThe paper uses Gaussian mixture models and full-batch gradient descent as the theoretical analysis objects.
50%UnverifiedK-means can be viewed as the hard-limit case of the EM algorithm on Gaussian mixture models, where the EM algorithm uses soft assignments while K-means uses hard assignments
50%