Indian buffet process

The Indian Buffet Process (IBP) is a Bayesian method used for inferring the model space. It is one of several sampling methods for the Dirichlet Process, along with the pólya urn, Chinese Restaurant Process, Hierarchical Dirichlet Process, and Stick Breaking process. The IBP is also used in the sparse CCA framework to automatically select the number of correlation components.From: A Concise Introduction to Machine Learning [2019], A Bayesian Monotonic Non-parametric Dose-Response Model [2021], Multi-Label Dimensionality Reduction [2019]

Feature Learning

View Chapter

Purchase Book

Published in A. C. Faul, A Concise Introduction to Machine Learning, 2019

A. C. Faul

This chapter concentrates on inferring the building blocks which are essential for an adequate model of the data. Neural networks are revisited and their many variations explained, especially how design choices such as activation (propagation functions) and error functions depend on the task at hand. Error backpropagation is derived in generality as the application of the chain rule when differentiating the different error functions. Autoencoders are introduced as very simple network architecture and illustrated with an extensive example showing how neural networks create a model space for the data. The relationships to other techniques are explored to gain an intuitive understanding of the mechanisms at work. The chapter concludes with a Bayesian treatment of inferring the model space, illustrated by the Indian Buffet Process.

A Bayesian Monotonic Non-parametric Dose-Response Model

View Article

Journal Information

Published in Human and Ecological Risk Assessment: An International Journal, 2021

Faten S. Alamri, Edward L. Boone, David J. Edwards

Recall that one of the constraints for the AM-Spline is that and for decreasing percent to control data. In order to achieve this constraint the Stick-Breaking prior is chosen as it is a Dirichlet process (DP) that has the same constraints as our model. Let then where the αj are parameters of the Dirichlet distribution. The Dirichlet Process has several sampling methods such as: pólya urn, Chinese Restaurant Process, Hierarchical Dirichlet Process, Indian Buffet Process and the Stick Breaking process (Frigyik et al. 2010). For more about Stick breaking process see Müller et al. (2015).

Indian buffet process

Explore chapters and articles related to this topic

Feature Learning

A Bayesian Monotonic Non-parametric Dose-Response Model