| | |
| | |
Stat |
Members: 3643 Articles: 2'487'895 Articles rated: 2609
28 March 2024 |
|
| | | |
|
Article overview
| |
|
Learning Markov Models from Closed Loop Data-sets | Jonathan Epperlein
; Robert Shorten
; Sergiy Zhuk
; | Date: |
20 Jun 2017 | Abstract: | Many practical problems involve developing prediction and optimization tools
to manage supply and demand issues. Almost always, these prediction tools are
created with a view to affecting behavioural change, thus creating a feedback
loop. Clearly, successful applications actuating behavioural change, affect the
original model underpinning the predictor, leading to an inconsistency. This
feedback loop is often not considered in standard so-called Big Data learning
techniques which rely upon machine learning/statistical learning machinery. The
objective of this paper is to develop mathematically sound tools for the design
of predictor feedback systems. For this purpose we use the framework of Hidden
Markov Models (HMMs). More specifically, we assume that we observe a time
series which is a path (output) of a Markov chain ${R}$ modulated by another
Markov chain ${S}$, i.e. the transition matrix of ${R}$ is unknown and depends
on the current realisation of the state of ${S}$. The transition matrix of the
latter is also unknown. In other words, at each time instant, ${S}$ selects a
transition matrix for ${R}$ within a given set which consist of a number of
known and uncertain matrices of a given dimension. The state of ${S}$, in turn,
depends on the current state of ${R}$ thus introducing a feed-back loop. We
propose a modification of classical Baum-Welsch algorithm, a variant of
Expectation-Maximization family of methods, which estimates the transition
matrices of ${S}$ and ${R}$. Experimental study shows that most of the time our
method is identifying the "true" transition matrices which were used to
generate the data. In all cases, the likelihoods of our estimates are at least
as good as the likelihoods of the "true" matrices. | Source: | arXiv, 1706.6359 | Services: | Forum | Review | PDF | Favorites |
|
|
No review found.
Did you like this article?
Note: answers to reviews or questions about the article must be posted in the forum section.
Authors are not allowed to review their own article. They can use the forum section.
browser claudebot
|
| |
|
|
|
| News, job offers and information for researchers and scientists:
| |