• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.9
• /
• pp.4317-4335
• /
• 2018

Semi-supervised multi-view latent space learning is gaining considerable popularity recently in many machine learning applications due to the high cost and difficulty to obtain the large amount of label information of data. Although some semi-supervised multi-view latent space learning methods have been presented, there is still much space for improvement: 1) How to learn latent discriminant intact feature representations by employing data of multiple views; 2) How to exploit the manifold structure of both labeled and unlabeled point in the learned latent intact space effectively. To address the above issues, we propose an approach called semi-supervised multi-view manifold discriminant intact space learning ($SM^2DIS$) for image classification in this paper. $SM^2DIS$ aims to seek a manifold discriminant intact space for data of different views by making use of both the discriminant information of labeled data and the manifold structure of both labeled and unlabeled data. Experimental results on MNIST, COIL-20, Multi-PIE, and Caltech-101 databases demonstrate the effectiveness and robustness of our proposed approach.

• Proceedings of the KSRS Conference
• /
• 2005.10a
• /
• pp.415-417
• /
• 2005

In order to retrieve the latent and sensible heat fluxes, high-resolution airborne imageries with visible, near infrared, and thermal infrared bands and ground-base meteorology measurements are utilized in this paper. The retrieval scheme is based on the balance of surface energy budget and momentum equations. There are three basic surface parameters including surface albedo $(\alpha)$, normalized difference vegetation index (NOVI) and surface kinetic temperature (TO). Lowtran 7 code is used to correct the atmosphere effect. The imageries were taken on 28 April and 5 May 2003. From the scattering plot of data set, we observed the extreme dry and wet pixels to derive the fitting of dry and wet controlled lines, respectively. Then the sensible heat and latent heat fluxes are derived from through a partitioning factor A. The retrieved latent and sensible heat fluxes are compared with in situ measurements, including eddy correlation and porometer measurements. It is shown that the retrieved fluxes from our scheme match with the measurements better than those derived from the S-SEBI model.

• Journal of Intelligence and Information Systems
• /
• v.27 no.3
• /
• pp.175-197
• /
• 2021

Recently, with the development of deep learning technology, research on unstructured data analysis is being actively conducted, and it is showing remarkable results in various fields such as classification, summary, and generation. Among various text analysis fields, text classification is the most widely used technology in academia and industry. Text classification includes binary class classification with one label among two classes, multi-class classification with one label among several classes, and multi-label classification with multiple labels among several classes. In particular, multi-label classification requires a different training method from binary class classification and multi-class classification because of the characteristic of having multiple labels. In addition, since the number of labels to be predicted increases as the number of labels and classes increases, there is a limitation in that performance improvement is difficult due to an increase in prediction difficulty. To overcome these limitations, (i) compressing the initially given high-dimensional label space into a low-dimensional latent label space, (ii) after performing training to predict the compressed label, (iii) restoring the predicted label to the high-dimensional original label space, research on label embedding is being actively conducted. Typical label embedding techniques include Principal Label Space Transformation (PLST), Multi-Label Classification via Boolean Matrix Decomposition (MLC-BMaD), and Bayesian Multi-Label Compressed Sensing (BML-CS). However, since these techniques consider only the linear relationship between labels or compress the labels by random transformation, it is difficult to understand the non-linear relationship between labels, so there is a limitation in that it is not possible to create a latent label space sufficiently containing the information of the original label. Recently, there have been increasing attempts to improve performance by applying deep learning technology to label embedding. Label embedding using an autoencoder, a deep learning model that is effective for data compression and restoration, is representative. However, the traditional autoencoder-based label embedding has a limitation in that a large amount of information loss occurs when compressing a high-dimensional label space having a myriad of classes into a low-dimensional latent label space. This can be found in the gradient loss problem that occurs in the backpropagation process of learning. To solve this problem, skip connection was devised, and by adding the input of the layer to the output to prevent gradient loss during backpropagation, efficient learning is possible even when the layer is deep. Skip connection is mainly used for image feature extraction in convolutional neural networks, but studies using skip connection in autoencoder or label embedding process are still lacking. Therefore, in this study, we propose an autoencoder-based label embedding methodology in which skip connections are added to each of the encoder and decoder to form a low-dimensional latent label space that reflects the information of the high-dimensional label space well. In addition, the proposed methodology was applied to actual paper keywords to derive the high-dimensional keyword label space and the low-dimensional latent label space. Using this, we conducted an experiment to predict the compressed keyword vector existing in the latent label space from the paper abstract and to evaluate the multi-label classification by restoring the predicted keyword vector back to the original label space. As a result, the accuracy, precision, recall, and F1 score used as performance indicators showed far superior performance in multi-label classification based on the proposed methodology compared to traditional multi-label classification methods. This can be seen that the low-dimensional latent label space derived through the proposed methodology well reflected the information of the high-dimensional label space, which ultimately led to the improvement of the performance of the multi-label classification itself. In addition, the utility of the proposed methodology was identified by comparing the performance of the proposed methodology according to the domain characteristics and the number of dimensions of the latent label space.

• Journal of architectural history
• /
• v.15 no.3
• /
• pp.119-135
• /
• 2006

This study based on the hypothesis which the spatial qualities in the Mies's early works are not extinct but potentially immanent in his latter works. In Mies's early works, destruction of outline, centrifugal extension of fluid space, and asymmetry are distinctly showed. These qualities probably revert to the indefiniteness of space. In Mies's latter works, however, these dynamic qualities are disappeared. Geometrically precise outline and exact grid structure represent universal space derived from zero-degree pure box. These qualities probably revert to tile definiteness of space, characterized by the unmovable emptiness. Although Mies works vary in external form, his expression technique of space reveals continually both the qualify of definitive and indefinitive space. For example, in the Museum for a Small City(1942) unbuilt project. elements defined by the perspective are fixed and static, but elements defined by the collages are floated and dynamic. The former reigns over the realized buildings of Mies, while the latter seems to be latent in terms of Schein which transcends reality. If we can penetrate this point, it's possible to read the other side of Mies' architectural works, distinct from both the canonized interpretation and the excessive criticism. Point is that later works of Mies must be understood as interplay of universal space appeared as phenomenon and flowing elements latent in. Architectural space of Mies keeps a distance with actual space through latent manner of being while preserves the empirical actuality It provides us with an occasion which appears only in an instant, in which even the ordinary things reveal its poetics.

• Proceedings of the KSRS Conference
• /
• v.2
• /
• pp.752-755
• /
• 2006

Evaportranspiration is an important factor in hydrology cycle. Traditionally, it is measured by using basin or empirical formula with meteorology data, while it does not represent the evaportranspiration over a regional area. With the advent of improved remote sensing technology, it becomes a surface parameter of research interest in the field of remote sensing. Airborne and satellite imagery are utilized in this study. The high resolution airborne images include visible, near infrared, and thermal infrared bands and the satellite images are acquired by MODIS. Surface heat fluxes such as latent heat flux and sensible heat flux are estimate by using airborne and satellite images with surface meteorological measurements. We develop a new method to estimate the evaportranspiration over the rice paddy. The surface heat fluxes are initialized with a surface energy balance concept and iterated for convergent solution with atmospheric correct functions associated with aerodynamic resistance of heat transport. Furthermore, we redistribute the total net energy into sensible heat and latent heat fluxes. The result reveals that radiation and evaporation controlled extremes can be properly decided with both airborne and satellite images. The correlation coefficient of latent heat flux and sensible heat flux with corresponding in situ observations are 0.66 and 0.76, respectively. The relative root mean squared errors (RMSEs) for latent heat flux and sensible heat flux are 97.81 $(W/m^2)$ and 124.33 $(W/m^2)$, respectively. It is also shown that the newly developed retrieval scheme performs well when it is tested by using MODIS date.

• The Korean Journal of Applied Statistics
• /
• v.30 no.4
• /
• pp.555-566
• /
• 2017

For multivariate datasets with large number of variables, classical dimensional reduction methods such as principal component analysis may not be effective for data visualization. The underlying reason is that the dimensionality of the space of variables is often larger than two or three, while the visualization to the human eye is most effective with two or three dimensions. This paper proposes a working procedure which first partitions the variables into several "latent" clusters, explores individual data subsets, and finally integrates findings. We use R pakacage "ClustOfVar" for partitioning variables around latent dimensions and the principal component biplot method to visualize within-cluster patterns. Additionally, we use the technique for embedding supplementary variables to figure out the relationships between within-cluster variables and outside variables.

• Proceedings of the KSRS Conference
• /
• 2008.10a
• /
• pp.317-319
• /
• 2008

Traditionally, it is measured by using basin or empirical formula with meteorology data, while it does not represent the evaportransporation over a regional area. With the advent of improved remote sensing technology, it becomes feasible to assess the ET over a regional scale. Firstly, the IMAGINE ATCOR atmospheric module is used to preprocess for the MODIS imagery. Then MODIS satellite images which have been corrected by radiation and geometry in conjunction with the in-situ surface meteorological measurement are used to estimate the surface heat fluxes such as soil heat flux, sensible heat flux, and latent heat flux. In addition, the correlation coefficient between the derived latent heat and the in-situ measurement is found to be over 0.76. In the future, we will continue to monitor the surface heat fluxes of paddy rice field in Chiayi area.

• Speech Sciences
• /
• v.11 no.4
• /
• pp.75-88
• /
• 2004

In this paper, we propose a new approach to sequential linear regression adaptation of continuous density hidden Markov models (CDHMMs) based on transformation space model (TSM). The proposed TSM which characterizes the a priori knowledge of the training speakers associated with maximum likelihood linear regression (MLLR) matrix parameters is effectively described in terms of the latent variable models. The TSM provides various sources of information such as the correlation information, the prior distribution, and the prior knowledge of the regression parameters that are very useful for rapid adaptation. The quasi-Bayes (QB) estimation algorithm is formulated to incrementally update the hyperparameters of the TSM and regression matrices simultaneously. Experimental results showed that the proposed TSM approach is better than that of the conventional quasi-Bayes linear regression (QBLR) algorithm for a small amount of adaptation data.

• Smart Media Journal
• /
• v.10 no.3
• /
• pp.9-22
• /
• 2021

Basically, machine learning models use input data to produce results. Sometimes, the input data is too complicated for the models to learn useful patterns. Therefore, feature engineering is a crucial data preprocessing step for constructing a proper feature set to improve the performance of such models. One of the most efficient methods for automating feature engineering is the autoencoder, which transforms the data from its original space into a latent space. However certain factors, including the datasets, the machine learning models, and the number of dimensions of the latent space (denoted by k), should be carefully considered when using the autoencoder. In this study, we design a framework to compare two data preprocessing approaches: with and without autoencoder and to observe the impact of these factors on autoencoder. We then conduct experiments using autoencoders with classifiers on popular datasets. The empirical results provide a perspective regarding the best suited autoencoder for these factors.

• Proceedings of the KSRS Conference
• /
• 2008.10a
• /
• pp.68-71
• /
• 2008

Our goal is to obtain a better scientific understanding how to define the nature and role of remotely sensed land surface parameters and energy fluxes in the heat island phenomena, and local and regional weather and climate. By using the MODIS visible and thermal imagery data and analyzing the surface energy flux images associated with the change of the landcover and landuse in study area, we will estimate and present how significant is the magnitude of the heat island heat effect and its relation with the surface parameters and the energy fluxes in Taiwan. To achieve our objective, we used the energy budget components such as net radiation, soil heat flux, sensible heat flux, and latent heat flux in the study area of interest derived form remotely sensed data to understand the island heat effect. The result shows that the water is the most important component to decrease the temperature, and the more the consumed net radiation to latent heat, the lower urban surface temperature.