Learning Hidden Markov Models with Structured Transition Dynamics

被引:0
|
作者
Ma, Simin [1 ]
Dehghanian, Amin [1 ]
Garcia, Gian-Gabriel [1 ]
Serban, Nicoleta [1 ]
机构
[1] Georgia Inst Technol, H Milton Stewart Sch Ind & Syst Engn, Atlanta 30332, GA USA
基金
美国国家卫生研究院;
关键词
expectation-maximization algorithm; hidden Markov model; convex optimization; accelerated gradient method; statistical learning; LIKELIHOOD RATIO TEST; PATIENTS PRICE; CONCUSSION; ALGORITHM; STATEMENT; SOLVERS; PRIVACY; NUMBER;
D O I
10.1287/ijoc.2022.0342
中图分类号
TP39 [计算机的应用];
学科分类号
081203 ; 0835 ;
摘要
The hidden Markov model (HMM) provides a natural framework for modeling the dynamic evolution of latent diseases. The unknown probability matrices of HMMs can be learned through the well-known Baum-Welch algorithm, a special case of the expectation-maximization algorithm. In many disease models, the probability matrices possess nontrivial properties that may be represented through a set of linear constraints. In these cases, the traditional Baum-Welch algorithm is no longer applicable because the maximization step cannot be solved by an explicit formula. In this paper, we propose a novel approach to efficiently solve the maximization step problem under linear constraints by providing a Lagrangian dual reformulation that we solve by an accelerated gradient method. The performance of this approach critically depends on devising a fast method to compute the gradient in each iteration. For this purpose, we employ dual decomposition and derive Karush-Kuhn-Tucker conditions to reduce our problem into a set of single variable equations, solved using a simple bisection method. We apply this method to a case study on sports-related concussion and provide an extensive numerical study using simulation. We show that our approach is in orders of magnitude computationally faster and more accurate than other alternative approaches. Moreover, compared with other methods, our approach is far less sensitive with respect to increases in problem size. Overall, our contribution lies in the advancement of accurately and efficiently handling HMM parameter estimation under linear constraints, which comprises a wide range of applications in disease modeling and beyond.
引用
收藏
页数:26
相关论文
共 50 条
  • [1] Phase transition for parameter learning of hidden Markov models
    Rau, Nikita
    Luecke, Joerg
    Hartmann, Alexander K.
    PHYSICAL REVIEW E, 2021, 104 (04)
  • [2] Learning interaction dynamics with coupled hidden Markov models
    Rezek, I
    Sykacek, P
    Roberts, SJ
    IEE PROCEEDINGS-SCIENCE MEASUREMENT AND TECHNOLOGY, 2000, 147 (06) : 345 - 350
  • [3] NONPARAMETRIC LEARNING FOR HIDDEN MARKOV MODELS WITH PREFERENTIAL ATTACHMENT DYNAMICS
    Hensley, Asher A.
    Djuric, Petar M.
    2017 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP), 2017, : 3854 - 3858
  • [4] Incremental Construction of Structured Hidden Markov Models
    Galassi, Ugo
    Giordana, Attilio
    Saitta, Lorenza
    20TH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, 2007, : 798 - 803
  • [5] Creditworthiness dynamics and Hidden Markov Models
    Quirini, L.
    Vannucci, L.
    JOURNAL OF THE OPERATIONAL RESEARCH SOCIETY, 2014, 65 (03) : 323 - 330
  • [6] Learning Hidden Markov Sparse Models
    Li, Lin
    Scaglione, Anna
    2013 INFORMATION THEORY AND APPLICATIONS WORKSHOP (ITA), 2013,
  • [7] Learning Scripts as Hidden Markov Models
    Orr, J. Walker
    Tadepalli, Prasad
    Doppa, Janardhan Rao
    Fern, Xiaoli
    Dietterich, Thomas G.
    PROCEEDINGS OF THE TWENTY-EIGHTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, 2014, : 1565 - 1571
  • [8] Learning discrete hidden Markov models
    Meloni, LGP
    COMPUTER APPLICATIONS IN ENGINEERING EDUCATION, 2000, 8 (3-4) : 141 - 149
  • [9] Learning Hidden Quantum Markov Models
    Srinivasan, Siddarth
    Gordon, Geoff
    Boots, Byron
    INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND STATISTICS, VOL 84, 2018, 84
  • [10] Online learning with hidden Markov models
    Mongillo, Gianluigi
    Deneve, Sophie
    NEURAL COMPUTATION, 2008, 20 (07) : 1706 - 1716