Maximum Likelihood Methods in Unknown Latent Variables: Difference between revisions

From Algorithm Wiki
Jump to navigation Jump to search
(Created page with "{{DISPLAYTITLE:Maximum Likelihood Methods in Unknown Latent Variables (Maximum Likelihood Methods in Unknown Latent Variables)}} == Description == In this problem, the goal is to compute maximum-likelihood estimates when the observations can be viewed as incomplete data. == Parameters == No parameters found. == Table of Algorithms == {| class="wikitable sortable" style="text-align:center;" width="100%" ! Name !! Year !! Time !! Space !! Approximation Factor !!...")
 
No edit summary
 
(3 intermediate revisions by the same user not shown)
Line 6: Line 6:
== Parameters ==  
== Parameters ==  


No parameters found.
$n$: number of observations in sample
 
$r$: number of parameters + latent variables


== Table of Algorithms ==  
== Table of Algorithms ==  
Line 18: Line 20:
| [[Expectation-Maximization (EM) algorithm (Maximum Likelihood Methods in Unknown Latent Variables Maximum Likelihood Methods in Unknown Latent Variables)|Expectation-Maximization (EM) algorithm]] || 1977 || $O(n^{3})$ || $O(n+r)$? || Exact || Deterministic || [https://www.jstor.org/stable/2984875 Time]
| [[Expectation-Maximization (EM) algorithm (Maximum Likelihood Methods in Unknown Latent Variables Maximum Likelihood Methods in Unknown Latent Variables)|Expectation-Maximization (EM) algorithm]] || 1977 || $O(n^{3})$ || $O(n+r)$? || Exact || Deterministic || [https://www.jstor.org/stable/2984875 Time]
|-
|-
| [[EM with Quasi-Newton Methods (Jamshidian; Mortaza; Jennrich; Robert I.) (Maximum Likelihood Methods in Unknown Latent Variables Maximum Likelihood Methods in Unknown Latent Variables)|EM with Quasi-Newton Methods (Jamshidian; Mortaza; Jennrich; Robert I.)]] || 1997 || $O(n^{2} log^{3} n)$ || $O(n+r^{2})$? || Exact || Deterministic || [https://rss.onlinelibrary.wiley.com/doi/abs/10.1111/1467-9868.00083 Time]
| [[EM with Quasi-Newton Methods (Jamshidian; Mortaza; Jennrich; Robert I.) (Maximum Likelihood Methods in Unknown Latent Variables Maximum Likelihood Methods in Unknown Latent Variables)|EM with Quasi-Newton Methods (Jamshidian; Mortaza; Jennrich; Robert I.)]] || 1997 || $O(n^{2} \log^{3} n)$ || $O(n+r^{2})$? || Exact || Deterministic || [https://rss.onlinelibrary.wiley.com/doi/abs/10.1111/1467-9868.00083 Time]
|-
|-
| [[Parameter-expanded expectation maximization (PX-EM) (Maximum Likelihood Methods in Unknown Latent Variables Maximum Likelihood Methods in Unknown Latent Variables)|Parameter-expanded expectation maximization (PX-EM)]] || 1998 || $O(n^{3})$ || $O(n+r)$? || Exact || Deterministic || [http://www.stat.ucla.edu/~ywu/research/papers/PXEM.pdf Time]
| [[Parameter-expanded expectation maximization (PX-EM) (Maximum Likelihood Methods in Unknown Latent Variables Maximum Likelihood Methods in Unknown Latent Variables)|Parameter-expanded expectation maximization (PX-EM)]] || 1998 || $O(n^{3})$ || $O(n+r)$? || Exact || Deterministic || [http://www.stat.ucla.edu/~ywu/research/papers/PXEM.pdf Time]
Line 27: Line 29:
|-
|-
| [[α-EM Algorithm (Maximum Likelihood Methods in Unknown Latent Variables Maximum Likelihood Methods in Unknown Latent Variables)|α-EM Algorithm]] || 2003 || $O(n^{3})$ || $O(n+r)$? || Exact || Deterministic || [https://waseda.pure.elsevier.com/en/publications/the-%CE%B1-em-algorithm-surrogate-likelihood-maximization-using-%CE%B1-loga Time]
| [[α-EM Algorithm (Maximum Likelihood Methods in Unknown Latent Variables Maximum Likelihood Methods in Unknown Latent Variables)|α-EM Algorithm]] || 2003 || $O(n^{3})$ || $O(n+r)$? || Exact || Deterministic || [https://waseda.pure.elsevier.com/en/publications/the-%CE%B1-em-algorithm-surrogate-likelihood-maximization-using-%CE%B1-loga Time]
|-
| [[Shaban; Amirreza; Mehrdad; Farajtabar (Maximum Likelihood Methods in Unknown Latent Variables; multi-view model, discrete observations Maximum Likelihood Methods in Unknown Latent Variables)|Shaban; Amirreza; Mehrdad; Farajtabar]] || 2015 || $O(n^{2} \log^{2} n)$ || $O(kd+d^{3})$?? || Exact || Deterministic || [https://faculty.cc.gatech.edu/~bboots3/files/SpectralExteriorPoint-NIPSWorkshop.pdf Time]
|-
| [[alpha-HMM (Matsuyama, Yasuo) (Maximum Likelihood Methods in Unknown Latent Variables, Hidden Markov Models Maximum Likelihood Methods in Unknown Latent Variables)|alpha-HMM (Matsuyama, Yasuo)]] || 2011 || $O(n^{2} \log^{2} n)$ ||  || Exact || Deterministic || [https://ieeexplore.ieee.org/abstract/document/7895145 Time]
|-
|-
|}
|}


== Time Complexity graph ==  
== Time Complexity Graph ==  


[[File:Maximum Likelihood Methods in Unknown Latent Variables - Time.png|1000px]]
[[File:Maximum Likelihood Methods in Unknown Latent Variables - Time.png|1000px]]
== Space Complexity graph ==
[[File:Maximum Likelihood Methods in Unknown Latent Variables - Space.png|1000px]]
== Pareto Decades graph ==
[[File:Maximum Likelihood Methods in Unknown Latent Variables - Pareto Frontier.png|1000px]]

Latest revision as of 09:10, 28 April 2023

Description

In this problem, the goal is to compute maximum-likelihood estimates when the observations can be viewed as incomplete data.

Parameters

$n$: number of observations in sample

$r$: number of parameters + latent variables

Table of Algorithms

Name Year Time Space Approximation Factor Model Reference
Expectation-Maximization (EM) algorithm 1977 $O(n^{3})$ $O(n+r)$? Exact Deterministic Time
EM with Quasi-Newton Methods (Jamshidian; Mortaza; Jennrich; Robert I.) 1997 $O(n^{2} \log^{3} n)$ $O(n+r^{2})$? Exact Deterministic Time
Parameter-expanded expectation maximization (PX-EM) 1998 $O(n^{3})$ $O(n+r)$? Exact Deterministic Time
Expectation conditional maximization (ECM) 1993 $O(n^{3})$ $O(n+r)$? Exact Deterministic Time
Expectation conditional maximization either (ECME) (Liu; Chuanhai; Rubin; Donald B) 1994 $O(n^{3})$ $O(n+r)$? Exact Deterministic Time
α-EM Algorithm 2003 $O(n^{3})$ $O(n+r)$? Exact Deterministic Time
Shaban; Amirreza; Mehrdad; Farajtabar 2015 $O(n^{2} \log^{2} n)$ $O(kd+d^{3})$?? Exact Deterministic Time
alpha-HMM (Matsuyama, Yasuo) 2011 $O(n^{2} \log^{2} n)$ Exact Deterministic Time

Time Complexity Graph

Maximum Likelihood Methods in Unknown Latent Variables - Time.png