2020/01/07 by Yohan Jung, Jinkyoo Park, Jung, Yohan +1 · 1 citation
Computer Science · #Anomaly Detection Techniques and Applications #FOS: Computer and information sciences #Gaussian Processes and Bayesian Inference #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Time Series Analysis and Forecasting
paper · pdf · doi:10.48550/arxiv.2001.01917
openalex publication_date 2020/01/07 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
Hidden Markov Model (HMM) combined with Gaussian Process (GP) emission can be\neffectively used to estimate the hidden state with a sequence of complex\ninput-output relational observations. Especially when the spectral mixture (SM)\nkernel is used for GP emission, we call this model as a hybrid HMM-GPSM. This\nmodel can effectively model the sequence of time-series data. However, because\nof a large number of parameters for the SM kernel, this model can not\neffectively be trained with a large volume of data having (1) long sequence for\nstate transition and 2) a large number of time-series dataset in each sequence.\nThis paper proposes a scalable learning method for HMM-GPSM. To effectively\ntrain the model with a long sequence, the proposed method employs a Stochastic\nVariational Inference (SVI) approach. Also, to effectively process a large\nnumber of data point each time-series data, we approximate the SM kernel using\nReparametrized Random Fourier Feature (R-RFF). The combination of these two\ntechniques significantly reduces the training time. We validate the proposed\nlearning method in terms of its hidden-sate estimation accuracy and computation\ntime using large-scale synthetic and real data sets with missing values.\n