• Communications for Statistical Applications and Methods
• /
• v.25 no.5
• /
• pp.453-470
• /
• 2018

We study how to distinguish the parameters of the sparse autoregressive (AR) process from zero using a non-convex penalized estimation. A class of non-convex penalties are considered that include the smoothly clipped absolute deviation and minimax concave penalties as special examples. We prove that the penalized estimators achieve some standard theoretical properties such as weak and strong oracle properties which have been proved in sparse linear regression framework. The results hold when the maximal order of the AR process increases to infinity and the minimal size of true non-zero parameters decreases toward zero as the sample size increases. Further, we construct a practical method to select tuning parameters using generalized information criterion, of which the minimizer asymptotically recovers the best theoretical non-penalized estimator of the sparse AR process. Simulation studies are given to confirm the theoretical results.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.26 no.6
• /
• pp.89-96
• /
• 1989

An improved method on texture coding, which is a part of the region growing based image coding, is presented in this paper. An image is segmented into stochastic regions which can be described as a stochastic random field, and non-stochastic ones in order to efficiently represent texture. In the texture coding and reconstruction, an autoregressive model is used for the stochastic regions, while a two-dimensional polynomial approximation is used for the non-stochastic ones. This proposed method leads to a better subjective quality, relatively higher compression ratio and shorter processing time for coding and reconstructing than the conventional method which uses only two-dimensional polynomial approximation.

• Journal of the Korean Statistical Society
• /
• v.30 no.1
• /
• pp.31-39
• /
• 2001

For estimating parameters of possibly nonlinear and/or non-stationary seasonal autoregressive(AR) processes, we introduce a new instrumental variable method which use the direction vector of the regressors in the same period as an instrument. On the basis of the new estimator, we propose new seasonal random walk tests whose limiting null distributions are standard normal regardless of the period of seasonality and types of mean adjustments. Monte-Carlo simulation shows that he powers of he proposed tests are better than those of the tests based on ordinary least squares estimator(OLSE).

• Communications for Statistical Applications and Methods
• /
• v.25 no.4
• /
• pp.431-442
• /
• 2018

This paper highlights the computational explosion issues in the autoregressive moving average approach of frequency estimation of sinusoidal data with a large sample size. A new algorithm is proposed to circumvent the computational explosion difficulty in the conditional least-square estimation method. Notice that sinusoidal pattern can be generated by a non-invertible non-stationary autoregressive moving average (ARMA) model. The computational explosion is shown to be closely related to the non-invertibility of the equivalent ARMA model. Simulation studies illustrate the computational explosion phenomenon and show that the proposed algorithm can efficiently overcome computational explosion difficulty. Real data example of sunspot number is provided to illustrate the application of the proposed algorithm to the time series data exhibiting sinusoidal pattern.

• Communications for Statistical Applications and Methods
• /
• v.24 no.1
• /
• pp.1-13
• /
• 2017

This manuscript summarized advances in bootstrap methods for long-range dependent time series data. The stationary linear long-memory process is briefly described, which is a target process for bootstrap methodologies on time-domain and frequency-domain in this review. We illustrate time-domain bootstrap under long-range dependence, moving or non-overlapping block bootstraps, and the autoregressive-sieve bootstrap. In particular, block bootstrap methodologies need an adjustment factor for the distribution estimation of the sample mean in contrast to applications to weak dependent time processes. However, the autoregressive-sieve bootstrap does not need any other modification for application to long-memory. The frequency domain bootstrap for Whittle estimation is provided using parametric spectral density estimates because there is no current nonparametric spectral density estimation method using a kernel function for the linear long-range dependent time process.

• The Korean Journal of Applied Statistics
• /
• v.30 no.3
• /
• pp.323-334
• /
• 2017

We note that the dynamics of realized volatilities (RVs) are near the boundary between stationarity and non-stationarity because RVs have persistent long-memory and are often subject to fairly large outlying values. To forecast realized volatility, we consider a new method that adaptively use models with and without unit root according to the abnormality of observed RV: heterogeneous autoregressive (HAR) model and the Integrated HAR (IHAR) model. The resulting method is called the IHAR-O-HAR method. In an out-of-sample forecast comparison for the realized volatility datasets of the 3 major indexes of the S&P 500, the NASDAQ, and the Nikkei 225, the new IHAR-O-HAR method is shown superior to the existing HAR and IHAR method.

• The Korean Journal of Applied Statistics
• /
• v.29 no.1
• /
• pp.27-39
• /
• 2016

This paper considers variable selection in the sparse vector autoregressive (sVAR) model where sparsity comes from setting small coefficients to exact zeros. In the estimation perspective, Davis et al. (2015) showed that the lasso type of regularization method is successful because it provides a simultaneous variable selection and parameter estimation even for time series data. However, their simulations study reports that the regular lasso overestimates the number of non-zero coefficients, hence its finite sample performance needs improvements. In this article, we show that the adaptive lasso significantly improves the performance where the adaptive lasso finds the sparsity patterns superior to the regular lasso. Some tuning parameter selections in the adaptive lasso are also discussed from the simulations study.

• The Journal of the Convergence on Culture Technology
• /
• v.7 no.1
• /
• pp.640-645
• /
• 2021

The deep learning based end-to-end TTS system consists of Text2Mel module that generates spectrogram from text, and vocoder module that synthesizes speech signals from spectrogram. Recently, by applying deep learning technology to the TTS system the intelligibility and naturalness of the synthesized speech is as improved as human vocalization. However, it has the disadvantage that the inference speed for synthesizing speech is very slow compared to the conventional method. The inference speed can be improved by applying the non-autoregressive method which can generate speech samples in parallel independent of previously generated samples. In this paper, we introduce FastSpeech, FastSpeech 2, and FastPitch as Text2Mel technology, and Parallel WaveGAN, Multi-band MelGAN, and WaveGlow as vocoder technology applying non-autoregressive method. And we implement them to verify whether it can be processed in real time. Experimental results show that by the obtained RTF all the presented methods are sufficiently capable of real-time processing. And it can be seen that the size of the learned model is about tens to hundreds of megabytes except WaveGlow, and it can be applied to the embedded environment where the memory is limited.

• The Korean Journal of Applied Statistics
• /
• v.27 no.6
• /
• pp.1039-1047
• /
• 2014

An autoregressive model with normal errors is a natural model that attempts to fit time series data. More flexible models that include normal distribution as a special case are necessary because they can cover normality to non-normality models. The skewed exponential power distribution is a possible candidate for autoregressive models errors that may have tails lighter(platykurtic) or heavier(leptokurtic) than normal and skewness; in addition, the use of skewed exponential power distribution can reduce the influence of outliers and consequently increases the robustness of the analysis. We use SIR algorithm and grid method for an efficient Bayesian estimation.

• Journal of Digital Convergence
• /
• v.10 no.9
• /
• pp.357-362
• /
• 2012

We have compared and predicted for non-linear time series data which are real data having different variences using GRCA(1) model and neural network method. In particular, using Korea Composite Stock Price Index rate, mean square errors of prediction are obtained in genaralized random coefficient autoregressive model and neural network method. Neural network method prove to be better in short-term forecasting, however GRCA(1) model perform well in long-term forecasting.