Hsiao-Tien Pao

A comparison of neural network and multiple regression analysis in modeling capital structure

Empirical studies of the variation in debt ratios across firms have used statistical models singularly to analyze the important determinants of capital structure. Researchers, however, rarely employ non-linear models to examine the determinants and make little effort to identify a superior prediction model. This study adopts multiple linear regressions and artificial neural networks (ANN) models with seven explanatory variables of corporations feature and three external macro-economic control variables to analyze the important determinants of capital structures of the high-tech and traditional industries in Taiwan, respectively. Results of this study show that the determinants of capital structure are different in both industries. The major different determinants are business-risk and growth opportunities. Based on the values of RMSE, the ANN models achieve a better fit and forecast than the regression models for debt ratio, and ANNs are cable of catching sophisticated non-linear integrating effects in both industries. It seems that the relationships between debt ratio and independent variables are not linear. Managers can apply these results for their dynamic adjustment of capital structure in achieving optimality and maximizing firms value.

An EM based multiple instance learning method for image classification

In this paper, we propose an EM based learning algorithm to provide a comprehensive procedure for maximizing the measurement of diverse density on given multiple Instances. Furthermore, the new EM based learning framework converts an MI problem into a single-instance treatment by using EM to maximize the instance responsibility for the corresponding label of each bag. To learn a desired image class, a user may select a set of exemplar images and label them to be conceptual related (positive) or conceptual unrelated (negative) images. A positive image consists of at least one object that the user may be interested, and a negative image should not contain any object that the user may be interested. By using the proposed EM based learning algorithm, an image retrieval prototype system is implemented. Experimental results show that for only a few times of relearning cycles, the prototype system can retrieve users favor images from WWW over Internet.

Adaptive watermarking using relationships between wavelet coefficients

We propose an approach that hides watermarks in relationships between wavelet coefficients. This approach minimizes the perceptual distortion of watermarked images, measured by PSNR or just noticeable distortion (JND). Therefore, the strength of watermarks can be enlarged to increase the robustness of the watermarks, while keeping the quality of watermarked images visually acceptable. Experimental results show that the watermark is still detectable after common image processing operations such as JPEG compression, Gaussian filtering and sharpening.

User adaptive handwriting recognition by self-growing probabilistic decision-based neural networks

It is generally agreed that, for a given handwriting recognition task, a user dependent system usually outperforms a user independent system, as long as a sufficient amount of training data is available. When the amount of user training data is limited, however, such a performance gain is not guaranteed. One way to improve the performance is to make use of existing knowledge, contained in a rich multiuser data base, so that a minimum amount of training data is sufficient to initialize a model for the new user.We mainly address the user adaption issues for a handwriting recognition system. Based on self-growing probabilistic decision-based neural networks (SPDNNs), user adaptation of the parameters of SPDNN is formulated as incremental reinforced and antireinforced learning procedures, which are easily integrated into the batched training procedures of the SPDNN. In this study, we developed 1) an SPDNN based handwriting recognition system; 2) a two-stage recognition structure; and 3) a three-phase training methodology for a) a global coarse classifier (stage 1); b) a user independent hand written character recognizer (stage 2); and c) a user adaptation module on a personal computer. With training and testing on a 600-word commonly used Chinese character set, the recognition results indicate that the user adaptation module significantly improved the recognition accuracy. The average recognition rate increased from 44.2% to 82.4% in five adapting cycles, and the performance could finally increase up to 90.2% in ten adapting cycles.

A self-growing probabilistic decision-based neural network with automatic data clustering

In this paper, we propose a new clustering algorithm for a mixture of Gaussian-based neural network and self-growing probabilistic decision-based neural networks (SPDNN). The proposed self-growing cluster learning (SGCL) algorithm is able to find the natural number of prototypes based on a self-growing validity measure, Bayesian information criterion (BIC). The learning process starts from a single prototype randomly initialized in the feature space and grows adaptively during the learning process until most appropriate number of prototypes are found. We have conducted numerical and real-world experiments to demonstrate the effectiveness of the SGCL algorithm. In the results of using SGCL to train the SPDNN for data clustering and speaker identification problems, we have observed a noticeable improvement among various model-based or vector quantization-based classification schemes.

A neural network approach to m-daily-ahead electricity price prediction

This paper proposes an artificial neural network (ANN) model to predict m-daily-ahead electricity price using direct forecasting approach on European Energy Exchange (EEX) market. The most important characteristic of this model is the single output node for m-period-ahead forecasts. The potentials of ANNs are investigated by employing cross-validation schemes. Out-of-sample performance evaluated with three criteria across five forecasting horizons shows that the proposed ANNs are more robust multi-step-ahead forecasting method than autoregressive error models (AUTOREG). Moreover, ANN predictions are quite accurate even when the length of forecast horizon is relatively short or long.

Constructing and application of multimedia TV-news archives

This paper addresses an integrated information mining techniques for broadcasting TV-news. This utilizes technique from the fields of acoustic, image, and video analysis, for information on news story title, newsman and scene identification. The goal is to construct a compact yet meaningful abstraction of broadcast TV-news, allowing users to browse through large amounts of data in a non-linear fashion with flexibility and efficiency. By adding acoustic analysis, a news program can be partitioned into news and commercial clips, with 90% accuracy on a data set of 400h TV-news recorded off the air from July 2005 to August 2006. By applying speaker identification and/or image detection techniques, each news stories can be segmented with a better accuracy of 95.92%. On-screen captions or subtitles are recognized by OCR techniques to produce the text title of each news stories. The extracted title words can be used to link or to navigate more related news contents on the WWW. In cooperation with facial and scene analysis and recognition techniques, OCR results can provide users with multimodal query on specific news stories. Some experimental results are presented and discussed for the system reliability, performance evaluation and comparison.

A self-growing probabilistic decision-based neural network for anchor/speaker identification

In this paper, we propose a new clustering algorithm for a mixture Gaussian based neural network, called Self-growing Probabilistic decision-based neural networks (SPDNN). The proposed Self-growing cluster learning (SGCL) algorithm is able to find the natural number of prototypes based on a self-growing validity measure, Bayesian Information Criterion (BIC). The learning process starts with a single prototype randomly initialized in the feature space and grows adaptively during the learning process until most appropriate number of prototypes are found. We have conduct numerical and real world experiments to demostrate the effectiveness of the SGCL algorithm. In the results of using SGCL to trainin the SPDNN for anchor/speaker identification, we have observed noticeable improvement among various model-based or vector quantization-based classification schemes.

Image classification and indexing by EM based multiple- instance learning

In this paper, we propose an EM based Multiple-Instance learning algorithm for the image classification and indexing. To learn a desired image class, a set of exemplar images are selected by a user. Each example is labeled as conceptual related (positive) or conceptual unrelated (negative) image. A positive image consists of at least one user interested object, and a negative example should not contain any user interested object. By using the proposed learning algorithm, an image classification system can learn the users preferred image class from the positive and negative examples. We have built a prototype system to retrieve user desired images. The experimental results show that for only a few times of relearning, a user can use the prototype system to retrieve favor images from the WWW over Internet.

Constructing and application of multimedia TV news archives

This paper addresses an integrated information mining techniques for broadcasting TV-news. The utilizes technique from the fields of acoustic, image, and video analysis, for information on news title, reporters and news background. The goal is to construct a compact yet meaningful abstraction of broadcast TV news, allowing users to browse through large amounts of data in a non-linear fashion with flexibility and efficiency. By using acoustic analysis, a news program can be partitioned into news and commercial clips, with 90% accuracy on a data set of 400 hours TV-news recorded off the air from July 2005 to August of 2006. By applying additional speaker identification and/or image detection techniques, each news stories can be segmented with a better accuracy of 95.92%. On screen captions and screen characters are recognized by video OCR techniques to produce the title of each news stories. Then keywords can be extracted from title to link related news contents on the WWW. In cooperation with facial and scene analysis and recognition techniques, OCR results can provide users with multimodal query on specific news stories.