Abstract
No abstract.
Decisions in the presence of latent variables
We correct errors in equations (14), (15) and (19) of the main text.
Equations (14) and (15)
Nature’s probability of flipping either coin does not actually depend on the agent’s prediction, so we can replace the conditional probabilities p0(θ|x) by p0(θ). We have then an inner variational problem:
with the solution
and the normalization constant and an outer variational problem as described by equation (16) in the main text. Note that deliberation renders the two variables x and θ dependent.
Equation (19)
In the case of α = β and uniform prior , equation (17) reduces to
where . Note that eαU(x,θ)/Z α is in general not a conditional distribution. However, equation (19) can be equivalently rewritten as
where we have expanded the fraction by .
This last equality can also be obtained by stating the same variational problem in reverse causal order of x and θ, which is the natural statement of the Thompson sampling problem. The nested variational problem then becomes
with the solutions
and
with normalization constant . In the limit α → ∞ and β → 0, the Thompson sampling agent is determined by the solutions and p(θ)=p0(θ). Sampling an action from is much cheaper than sampling an action from equation (18) because of the reversed causal order in θ and x, which implies that β/α→ 0 in equation (ii) instead of β/α→∞ as in equation (17).
In the case of α=β the solutions for the two different causal orders of x and θ are equivalent. Assuming again a uniform prior , we can compute the Thompson sampling agent from equation (i) and equation (ii) for α=β to be
which is exactly equivalent to p(x) in equation (19). To sample from equation (19), we draw θ~p0(θ) and accept if u≤eαU(x,θ)/eαT, where .
References
Ortega, PA, Braun DA: Generalized Thompson sampling for sequential decision-making and causal inference. Complex Adaptive Systems Modeling 2014, 2: 2.
Author information
Authors and Affiliations
Corresponding authors
Additional information
The online version of the original article can be found at 10.1186/2194-3206-2-2
Rights and permissions
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0), which permits use, duplication, adaptation, distribution, and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
About this article
Cite this article
Ortega, P.A., Braun, D.A. Erratum to: Generalized Thompson sampling for sequential decision-making and causal inference. Complex Adapt Syst Model 2, 4 (2014). https://doi.org/10.1186/s40294-014-0004-x
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s40294-014-0004-x