Hidden Markov models are tractable to capture long-term dependencies but intractable to compute the transition probabilities of higher-order process. We propose a neural hidden Markov models to compute the transition probabilities of higher-order hidden Markov model by a neural network and reduce the cost of computation. It is applied for time-aware recommender systems to show the benefits from the hybrid of combining neural network and hidden Markov model. We implement the recommender system and experiment on real datasets to demonstrate better performances over the existing recommender systems.
CITATION STYLE
Lin, Z., & Song, J. (2019). Neural Hidden Markov Model. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 11978 LNAI, pp. 37–54). Springer. https://doi.org/10.1007/978-3-030-37494-5_3
Mendeley helps you to discover research relevant for your work.