2020/06/04 by Matt Baucum, Baucum, Matt, Anahita Khojandi +3 · 1 citation
Computer Science · Mathematics · #Applications (stat.AP) #Bayesian Methods and Mixture Models #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Machine Learning in Healthcare #Methodology (stat.ME) #Statistical Methods and Inference
paper · pdf · doi:10.48550/arxiv.2006.03151
openalex publication_date 2020/06/04 · openalex created_date 2022/10/01 · openalex updated_date 2026/07/28
Hidden Markov models (HMMs) are commonly used for disease progression\nmodeling when the true patient health state is not fully known. Since HMMs\ntypically have multiple local optima, incorporating additional patient\ncovariates can improve parameter estimation and predictive performance. To\nallow for this, we develop hidden Markov recurrent neural networks (HMRNNs), a\nspecial case of recurrent neural networks that combine neural networks'\nflexibility with HMMs' interpretability. The HMRNN can be reduced to a standard\nHMM, with an identical likelihood function and parameter interpretations, but\nit can also combine an HMM with other predictive neural networks that take\npatient information as input. The HMRNN estimates all parameters simultaneously\nvia gradient descent. Using a dataset of Alzheimer's disease patients, we\ndemonstrate how the HMRNN can combine an HMM with other predictive neural\nnetworks to improve disease forecasting and to offer a novel clinical\ninterpretation compared with a standard HMM trained via\nexpectation-maximization.\n