Dirichlet Process
Updated 2026-08-13
INTRODUCTION
English translation pending.
CORE DEFINITION
A stochastic process whose draws are themselves probability distributions, introduced as a prior for infinite mixture models. Because it places mass on distributions with countably many atoms, it allows the number of clusters to grow with the data rather than being specified beforehand. The qualification is that the concentration parameter directly controls how finely the data is partitioned, so it functions as a real modeling decision rather than a nuisance setting. Its marginal representation is the Chinese restaurant process and its constructive form is stick-breaking.
SCAFFOLDING EFFECT
Reduce cognitive load
- Cluster count relief: use it when you cannot justify fixing the number of groups in advance. - Rich-get-richer read: interpret new observations as likely joining existing clusters rather than forming new ones. - Granularity tuning: adjust the concentration parameter to trade coarse partitions against fine ones.
Anchor fast decisions
The prior assigns probability to partitions rather than to a fixed set of clusters, so observing data can only reweight the space of partitions instead of forcing observations into preset groups. The rich-get-richer property means existing clusters attract new data in proportion to their size, while the concentration parameter controls how often an entirely new cluster appears, so complexity grows only when the data demands it.
MINIMUM ACTION
In progress 0/1Practice this model in one real situation:
account_treeGenealogyexpand_more
menu_bookReferencesexpand_more
Source support: Explicit
- en.wikipedia.orghttps://en.wikipedia.org/wiki/Dirichlet_processverified
PRIVATE NOTES · Only visible to you
SAVED Q&A
ENTRY Q&A · Private saving available
Ask with a clear boundary
thinkingmodels answers from published entry context only.
Your question is sent to thinkingmodels. The answer uses public entry context only.
RELATED MODELS