Chinese restaurant process

An exchangeable partition process in which each arriving item joins an existing block in proportion to its size or starts a new block with parameter-controlled probability.

Version

v2 · 2026-09-19 · History

Domain-specific #

3673

Domain group

Formal Sciences

Origin domain

Experimental Design & Statistics

Subdomain

Bayesian Nonparametrics → Experimental Design & Statistics

Core Idea

The restaurant metaphor encodes a probability law over partitions; concentration and discount parameters select Dirichlet-process or Pitman–Yor variants and table labels have no intrinsic order. Sequential reinforcement makes occupied blocks more likely while retaining innovation probability, and marginalizing arrival order yields an exchangeable partition distribution. The abstraction is therefore identified by a declared carrier, a transformation or constraint over that carrier, and an invariant that tells an analyst whether the named structure is genuinely present.

The load-bearing residual is not the broad topic of bayesian nonparametrics. It is the domain-specific identity determined by the customer index set, table partition and labeling convention, concentration and discount parameters, seating probabilities, exchangeability, induced partition law and limiting claims are explicit.

How would you explain it like I'm…

Busy Table Game

Picture a restaurant with lots of round tables. When a new person comes in, they usually sit at a table that already has people, and busier tables are more tempting. But sometimes they start a brand-new table. After everyone arrives, you look at who ended up sitting together. The tables have no special order; only the groups matter.

The Restaurant Grouping Game

The Chinese restaurant process is a way of randomly splitting things into groups, told as a story. Customers enter one at a time. Each new customer either joins a table that already has people — more likely if it's a crowded table — or starts a new table, with a chance set by a number called the concentration. The tables are just groups; their names or order don't matter. A neat fact is that the chance of ending up with a particular grouping doesn't depend on the order people arrived.

Exchangeable Random Partition Process

The Chinese restaurant process is a probability rule for randomly dividing items into groups, told as a story about customers choosing tables. Customers arrive one at a time; each joins an occupied table with probability that grows with the number of people already there, or starts a new table with a probability set by a concentration parameter. This creates a rich-get-richer pattern, while always keeping some chance of a new group. The tables have no built-in order or labels; the result is really a partition, meaning just which customers share a table. Remarkably, the probability of a final grouping doesn't depend on the arrival order, a property called exchangeability. Adding a discount parameter gives a variant related to the Pitman–Yor process, while the basic version corresponds to the Dirichlet process.

 

The Chinese restaurant process is a sequential construction of an exchangeable probability law over partitions of the customer index set. With concentration α (and, in the two-parameter version, discount d), customer n+1 joins an existing table k with probability proportional to (n_k − d) and opens a new table with probability proportional to (α + K·d), where K is the current number of tables; d = 0 gives the Dirichlet-process CRP and 0 ≤ d < 1 the Pitman–Yor variant. Sequential reinforcement favors occupied blocks while preserving innovation probability. Although defined sequentially, the probability of the resulting partition is invariant to arrival order, so the partition distribution is exchangeable, and table labels carry no intrinsic order. It is the partition structure induced by sampling from a Dirichlet or Pitman–Yor process and underlies nonparametric mixture models. Precise use requires declaring the index set, labeling convention, parameters, seating rule, and any asymptotic claims (such as growth of the number of tables).

Scope of Application

Chinese restaurant process belongs to bayesian nonparametrics and is useful where the analyst can specify the typed bayesian nonparametrics carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets, then evaluate the customer index set, table partition and labeling convention, concentration and discount parameters, seating probabilities, exchangeability, induced partition law and limiting claims are explicit. The scope is broad within that domain but bounded by the need for the customer index set, table partition and labeling convention, concentration and discount parameters, seating probabilities, exchangeability, induced partition law and limiting claims are explicit. The entry records a descriptive analytical identity; practical use requires the governing domain's evidence, standards, and safety obligations.

Clarity

The abstraction clarifies a crowded vocabulary by making the customer index set, table partition and labeling convention, concentration and discount parameters, seating probabilities, exchangeability, induced partition law and limiting claims are explicit the center of the account. A claim should name the carrier, the governing operation or relation, the applicable assumptions, and the recognition test. A bare label is insufficient because the name Chinese restaurant process can be used for a formal identity, an implementation, or a neighboring result unless carrier and convention are stated.

Manages Complexity

Without the abstraction, an analyst must reason directly over many local details: the carrier roles, admissibility assumptions, competing conventions, derived invariants, boundary cases, and proof or validation obligations specific to Chinese restaurant process. Chinese restaurant process compresses them into the roles in the structural signature. That compression permits comparison across instances without erasing the variables that determine validity. It also exposes which details may be varied safely and which are constitutive.

Abstract Reasoning

1. Identify the carrier. State what the elements, states, objects, or observations are: the typed bayesian nonparametrics carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets. Reject examples whose alleged carrier belongs to a different problem.
2. Lock the constitutive rule. Express the customer index set, table partition and labeling convention, concentration and discount parameters, seating probabilities, exchangeability, induced partition law and limiting claims are explicit independently of one notation or implementation.

Knowledge Transfer

Knowledge transfers strongly among subfields of bayesian nonparametrics because they reuse the typed bayesian nonparametrics carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets, Sequential reinforcement makes occupied blocks more likely while retaining innovation probability, and marginalizing arrival order yields an exchangeable partition distribution., and type the carrier, state every parameter and convention in the definition, test that the customer index set, table partition and labeling convention, concentration and discount parameters, seating probabilities, exchangeability, induced partition law and limiting claims are explicit, compare the nearest accepted identity, and report counterexamples, uncertainty, and limiting cases.

Relationships to Other Abstractions

Current abstraction Chinese restaurant process Domain-specific

Parents (1) — outgoing typed relations

• Chinese restaurant process is a kind of Stochastic Process Prime

  It remains its own entry because its identity is fixed by the customer index set, table partition and labeling convention, concentration and discount parameters, seating probabilities, exchangeability, induced partition law and limiting claims.

Hierarchy path (1) — routes to 1 parentless root

• Chinese restaurant process → is a kind of Stochastic Process

Neighborhood in Abstraction Space

Chinese restaurant process sits in a crowded region of the domain-specific corpus (36^th percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Process, Data & Business Modeling (9 abstractions)

Nearest neighbors

• Exchangeable random variables — 0.90
• Quantile — 0.88
• Bayesian linear regression — 0.87
• Sunrise problem — 0.87
• Normal-inverse-gamma distribution — 0.87

Computed from structural-signature embeddings · 2026-10-25