Skip to main content

Questions tagged [dirichlet-process]

A family of stochastic processes whose realizations are probability distributions

Filter by
Sorted by
Tagged with
0 votes
0 answers
11 views

Parallel Tempering and Bayesian Non-Parametrics

In parallel tempering, we run multiple MCMC chains in an ascending temperature ladder, where the posterior density of the $i$th chain is exponentiated to the reciprocal of the temperature of the $i$th ...
Faydey's user avatar
  • 225
0 votes
0 answers
28 views

Bayesian Hierarchical Clustering prior update

I am working through Heller and Ghahramani's "Bayesian Hierarchical Clustering" paper (https://www2.stat.duke.edu/~kheller/bhc.pdf) and things aren't quite working out the way I expect with ...
dataphile8's user avatar
0 votes
0 answers
14 views

Inference on latent variable with observation of its convolution with itself

Problem I have an inference problem where the data observed are univariate random numbers whose distribution is obtained as follows. A latent random variable X is first sampled from a parametric ...
Riccardo Buscicchio's user avatar
4 votes
2 answers
607 views

Dirichlet distribution with correlated components?

I am working with models that use Dirichlet distributions. However, I want to account for correlations between components. If this question is a duplicate, I'd also appreciate any pointers to the ...
Marco Pastor Mayo's user avatar
1 vote
0 answers
130 views

Stick-breaking construction of Dirichlet distribution vs Dirichlet process

Let $F_0$ be some probability measure and $\alpha > 0$ be the concentration parameter. I can draw a random distribution from $F\sim \mathrm{DP}(\alpha, F_0)$ using the stick-breaking construction: \...
Paweł Czyż's user avatar
0 votes
0 answers
83 views

Expected value (and variance) of a Dirichlet Process

Suppose I have a measure $G$ that follows a Dirichlet Process, $$G \sim DP(H_0,\alpha)$$ where $H_0$ is some base measure. Is there a closed form solution for the expected value of $G$?
dogs4ever's user avatar
0 votes
0 answers
17 views

If a scalar is added to a Dirichlet process with a normal base distribution, is the resulting process also Dirichlet?

Let $\theta_i | G \sim G$ and $G|\alpha \sim DP(\alpha, G_0)$ with $G_0 \sim N(0,\sigma^2)$. For a scalar k, does $\theta_i + k | G' \sim G'$ with $G'|\alpha \sim DP(\alpha, G'_0)$ and $G'_0 \sim N(k,\...
user128949's user avatar
1 vote
1 answer
163 views

Changing scan order of Gibbs Sampler on each iteration

I'm implementing an algorithm that requires the use of Gibbs Sampling and, due to the nature of the way I store the values, it would be efficient to change the order of the updates on each component ...
Edovt's user avatar
  • 31
1 vote
0 answers
65 views

Dirichlet Process posterior with partially observed data

Suppose I dipose of a set of independant observed couples $(x_1,y_1),...,(x_N, y_N)$ from a joint distribution $P(x,y)$. Furthermore, I suppose that the random distribution $P$ as a Dirichlet prior $P\...
Elouan's user avatar
  • 11
2 votes
0 answers
134 views

Stick-breaking: break sticks of decreasing lengths

The stick-breaking construction used for Dirichlet Processes can create an infinite sequence of probabilities $ \boldsymbol{\pi} $ (stick lengths) that sum to 1 via the following formulae: $\nu_i \sim ...
fm361's user avatar
  • 103
2 votes
1 answer
174 views

Understanding the blocked gibbs sampler for Dirichlet process

I've always implemented DP MM's using chinese restaurant process, which necessitates sequential sampling of cluster assignments (as cluster weights depend on current number of observations in cluster, ...
Faydey's user avatar
  • 225
2 votes
1 answer
71 views

Let $H$ be the base distribution of a Dirichlet process. How is this process well-defined in case $H(B_1) = 0$?

I have read that the parameters of Dirichlet distribution must be strictly positive. The Dirichlet distribution of order $K \geq 2$ with parameters $\alpha_{1}, \ldots, \alpha_{K} \color{blue}{> 0}...
Akira's user avatar
  • 433
0 votes
0 answers
312 views

Estimating the likelihood of a Dirichlet process

I am not sure if what I'm trying to achieve makes sense or is even possible, but I'd like to do MLE on a Dirichlet process mixture model. My reasoning is the following: If we can write out the ...
Pavlin's user avatar
  • 303
0 votes
0 answers
33 views

Can we use Dirichlet process to simultaneously estimate the number of mixtures and component distribution of a Bernoulli mixture?

Suppose I have a random sample on a Bernoulli random variable $\{X_i\}_{i=1}^N$ generated from model $p=\sum_{k=1}^K\pi_kp_k$,where $p\equiv Pr(X=1)$ and $p_k\equiv Pr(X=1|k)$, and $\pi_k$ are the ...
ExcitedSnail's user avatar
  • 2,914
0 votes
0 answers
95 views

Mixtures of Dirichlet multivariates or Dirichlet processes

I am exploring the properties of Dirichlet distributions and their parameters. When mixing two Dirichlet distributed random bivariates $$\mathbf{X}\equiv(X_1,X_2)\sim\text{Dir}(\alpha_1,\alpha_2)$$ ...
Riccardo Buscicchio's user avatar
0 votes
0 answers
31 views

Scoring rules for Dirichlet process mixtures

Let's say I do a sample-based fitting of a Dirichlet process model: $$ \begin{aligned} X_i &\sim f(x_i\mid \theta_i)\\ \theta_i &\sim G\\ G &\sim \text{DP}(\alpha, G_0) \end{aligned} $$ ...
Faydey's user avatar
  • 225
2 votes
0 answers
129 views

Marginal distributions of the Indian Buffet Process

In the construction of the Indian Buffet Process we have that customer $n_1$ chooses $\mbox{Poisson}\left(\frac{\alpha}{1}\right)$ dishes, $n_2$ chooses $\mbox{Poisson}\left(\frac{\alpha}{2}\right) \...
BadBayesian's user avatar
1 vote
0 answers
92 views

Online Stochastic Variational Inference for Dirichlet Process Mixture Models

There's a 2013 NeurIPS paper I'm trying to understand, Online Learning of Nonparametric Mixture Models via Sequential Variational Approximation. I have a few questions: Equation 2, which defines a ...
Rylan Schaeffer's user avatar
2 votes
0 answers
984 views

Dirichlet Process vs Gaussian Process?

I'm studying the Dirichlet Process (DP) and looking to the Gaussian Process (GP), which I've had more experience with, to help make connections. The GP can receive $X_{train}$, $X_{test}$, and $Y_{...
jbuddy_13's user avatar
  • 3,298
3 votes
1 answer
374 views

Why does Chinese Restaurant Table Distribution look like a Gaussian Distribution?

The Chinese Restaurant Table Distribution describes the probability distribution for the number of non-empty tables in the Chinese Restaurant Process after $T$ customers have been seated. Specifically,...
Rylan Schaeffer's user avatar
1 vote
1 answer
310 views

Validity of BIC for Dirichlet process mixture models

I am implementing clustering using Dirichlet process mixture models via scikit learn's Variational Bayesian Gaussian Mixture model. I arrived at the appropriate priors iteratively, and I am able to ...
nikarj's user avatar
  • 63
2 votes
1 answer
518 views

Chinese Restaurant Process: Expected cardinality (number of customers) of each block (table)?

Short version of the question: The Chinese Restaurant Process defines a distribution over partitions of $[T] := \{1, ...., T\}$. What is the expected cardinality of the $t$th block, where $t \in \{1, ....
Rylan Schaeffer's user avatar
3 votes
0 answers
131 views

A clarification in the original Dirichlet Process paper by Ferguson

I am reading the paper "Bayesian Analysis of Some Nonparametric Problems" by Ferguson where the Dirichlet process is introduced. There is a proposition 5 where the joint distribution of ...
honeybadger's user avatar
  • 1,562
1 vote
0 answers
47 views

Dirichlet Process vs Hierarchical Dirichlet Process: coupling among transitions on infinite HMM

I'm new to nonparametric Bayesian, and I am reading a paper about beam sampling for the infinite hidden Markov model. In the paper, it is mentioned that since there is no coupling among the ...
bubu's user avatar
  • 11
2 votes
0 answers
41 views

Unexpected zero on posterior density of Dirichlet process mixture

I was reading this notebook from the PyMC3 documentation about Dirichlet Process Mixtures and, on the last figure, the estimated density reaches almost zero for a particular value, despite the ...
PedroSebe's user avatar
  • 2,680
4 votes
1 answer
822 views

Implementation of a blocked Gibbs sampler for a mixture model with a Dirichlet-process prior

I am trying to understand and implement the blocked Gibbs sampler described on page 552 in Bayesian Data Analysis by Gelman et al. in the context of using a Dirichlet process as a prior in a mixture ...
Ivan's user avatar
  • 578
1 vote
0 answers
47 views

Marginal Cluster assignments for Dirichlet Process mixture model

I am watching Tamara' Broderick video on Dirichlet Process mixture models where she talks about computing $p(z_n = k | z_1,z_2,..z_{n-1})$ at ardoun 16:06. The z's are drawn from $$\rho_1 \sim beta(...
calveeen's user avatar
  • 1,116
1 vote
1 answer
53 views

Clarifying Wikipedia's Notation for Definition of a Dirichlet Process

Wikipedia gives the following definition of a Dirichlet process: What does the notation $H(B_i)$ and $X(B_i)$ mean? What does the notation $(X(B_1), ..., X(B_n))$ mean? My guess to the first ...
Rylan Schaeffer's user avatar
1 vote
1 answer
175 views

What is the Dirichlet Proces Mixture Models posterior

I am trying to understand Dirichlet Process Mixture models. One of the videos I have been watching is by Tamara Broderick. I think it is a very good introductory video to Dirichlet Process mixture ...
calveeen's user avatar
  • 1,116
1 vote
0 answers
53 views

Question about conditional probability in Dirichlet Process

Let $DP(\alpha_0, G_0)$ be a dirichlet process from which we sampled a distribution $G$. Furthermore we sample $\theta_1, ..., \theta_n$ from $G$ and let $A_1, ..., A_m$ be a partition on the support ...
Sebastian's user avatar
  • 3,094
3 votes
1 answer
80 views

Why are discrete random measures not dominated?

A statement I’ve seen without proof in many books and papers is that random probability measures obtained by normalizing a completely random measure, such as the Dirichlet process, are not dominated ...
B95's user avatar
  • 33
0 votes
2 answers
340 views

Fitting/Inference for Dirichlet Process/CRP for clustering

Please excuse my ignorance on this topic, I don't have much experience in nonparametric Bayes. I read about Dirichlet process clustering and the Chinese Restaurant Process analogies. I think I ...
bayesianpower's user avatar
0 votes
0 answers
47 views

Exchangabilty in CRF generative model for HDP

In the HDP Setting, the groups (or documents) are assumed exchangeable between them, and the samples (or words) within each group/topic are also exchangeable. Note the Setting chapter here However, ...
Dinari's user avatar
  • 151
4 votes
0 answers
324 views

Graphical model of the Gaussian mixture: where is n?

TL;DR: Where are the occupation numbers in the Graphical model of the GMM? I am implementing a Finite (to be adapted to infinite later) Gaussian Mixture Model. I am using the Gibbs sampler-ready ...
Lucidnonsense's user avatar
0 votes
1 answer
160 views

Understanding the sum in the stick breaking process

I am writing some R code for sampling from a Dirichlet process, but am having a hard time understanding how to take the final sum when there is a dirac delta ...
John Smith's user avatar
7 votes
1 answer
909 views

Kernel density estimate vs Dirichlet process mixture

Nowadays the Dirichlet process mixture (DPM) seems to be the default Bayesian approach for density estimation. My question is why not simply use the kernel density estimate (KDE) to model the density? ...
Frank's user avatar
  • 71
1 vote
0 answers
31 views

Is data modeled by dirichlet process mixture exchangeable?

Consider DPM model: $$ \begin{aligned} X_{i} | \phi_{i} & \sim F\left(x;\phi_{i}\right) \\ \phi_{1}, \phi_{2}, \cdots | & P \stackrel{iid}{\sim} P \\ P & \sim D P(\alpha G_0) \end{aligned} ...
Spaceship222's user avatar
1 vote
1 answer
403 views

Why is Algorithm 8 of Neal (2000) a valid sampler?

I have been having difficulty understanding why Algorithm 8 of Neal (2000) is a valid sampler. I am looking for lecture notes that include a nice explanation of the proof. Does anyone know of any ...
Tomislav's user avatar
  • 123
1 vote
0 answers
172 views

Dirichlet Process Concentration Parameter - collapse at zero

Background I've implemented the blocked gibbs sampler for sampling from the posterior of a dirichlet process mixture model as described on p.552 of Bayesian Data Analysis, placing a Gamma prior on the ...
Adam Peterson's user avatar
1 vote
1 answer
248 views

Bayesian Inference: Prior in Chinese Restaurant Process

For the Chinese restaurant process, as used in Dirichlet Process mixture models, we have a prior that data point i belongs to cluster j, where c is an indicator. n represents the total number of data ...
MJon10's user avatar
  • 13
3 votes
1 answer
585 views

Definition of the dirichlet process: what is the sequence of random variables

Reference material by Dr. Teh Definition Given a measureable set S, a base probability distribution H and a positive real number $\alpha$, the Dirichlet process $DP(H, \alpha)$ is a stochastic ...
fool's user avatar
  • 2,480
1 vote
1 answer
488 views

Clusters keep switching in Gibbs sampling of Dirichlet Process Mixture Model

All the code and data for this question is on GitHub (stackexchange.R script). I've got multivariate Bernoulli data that I'd like to analyse using Bayesian Mixture ...
Stuart Lacy's user avatar
5 votes
1 answer
216 views

Dirichlet process mixture modelling for a Gaussian likelihood

Let $\mathcal{Y} = (\mathbf{y}_1, \dots, \mathbf{y}_N)$ be data observed, such that each $\mathbf{y}_i \in \mathbb{R}^2$. Now conditional on unobserved cluster centres (means) $\mathcal{X} = (\mathbf{...
user202654's user avatar
2 votes
1 answer
285 views

Clarifying Dirichlet Process Mixture Probability Terms

Suppose I have a Dirichlet Process Mixture model defined as follows: $\alpha \sim G(a,b)\\ \pi|\alpha \sim \text{Dir}(\alpha)\\ z|\pi \sim \text{Cat}(\pi)\\ $ where $G$ is just a standard Gamma ...
tisPrimeTime's user avatar
0 votes
2 answers
2k views

Coding a simple Stick-Breaking Process in Python

I've just red the great 2012 blog post of Edwin Chen about Dirichlet Process with companion code in R and Ruby. Then I'm trying to translate the Stick-Breaking Process from R to Python. I've got this ...
Claude COULOMBE's user avatar
1 vote
1 answer
383 views

Gibbs sampler for Dirichlet Process concentration parameter

I am trying to implement a Gibbs sampler for Hierarchical Dirichlet process, but I cannot seem to correctly estimate the concentration parameters. I therefore started testing just this part of a ...
yassem's user avatar
  • 153
1 vote
0 answers
89 views

HDP: Gibbs sampler implementation

I am trying to recreate the model proposed by Gao et al. (2011), based on the Hierarchical Dirichlet Process proposed by Teh and al. (2005). To estimate the model (let's call it iHDP) I need to ...
yassem's user avatar
  • 153
1 vote
0 answers
356 views

Dirichlet-Multinomial Distribution with many zero counts

Short version: Is there someway to make the dirchlet-multinomial distribution sensitive to the presence of zero counts? Long version: I am attempting model metric positions in musical data. You can ...
user3474009's user avatar
2 votes
0 answers
137 views

How to interpret graphical model for Dirichlet process mixture for variational inference?

I am working through this paper by Blei and Jordan, which introduces variational inference for Dirichlet process mixtures. They derive an evidence lower bound (ELBO) function based on a stick breaking ...
scherm's user avatar
  • 1,035
2 votes
0 answers
78 views

How to actually draw a distribution from a given Dirichlet process?

I am wondering if there's an algorithm to actually draw a distribution from a given Dirichlet process. The closest thing I've came across is the stick-breaking construction of a Dirichlet process, but ...
Tetheras's user avatar