• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.8
• /
• pp.341-354
• /
• 2023

The title of a document is the brief summarization of the document. Readers can easily understand a document if we provide them with its title in their preferred styles and the languages. In this research, we propose a cross-lingual and style-based title generation model using multiple adapters. To train the model, we need a parallel corpus in several languages with different styles. It is quite difficult to construct this kind of parallel corpus; however, a monolingual title generation corpus of the same style can be built easily. Therefore, we apply a zero-shot strategy to generate a title in a different language and with a different style for an input document. A baseline model is Transformer consisting of an encoder and a decoder, pre-trained by several languages. The model is then equipped with multiple adapters for translation, languages, and styles. After the model learns a translation task from parallel corpus, it learns a title generation task from monolingual title generation corpus. When training the model with a task, we only activate an adapter that corresponds to the task. When generating a cross-lingual and style-based title, we only activate adapters that correspond to a target language and a target style. An experimental result shows that our proposed model is only as good as a pipeline model that first translates into a target language and then generates a title. There have been significant changes in natural language generation due to the emergence of large-scale language models. However, research to improve the performance of natural language generation using limited resources and limited data needs to continue. In this regard, this study seeks to explore the significance of such research.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.12
• /
• pp.519-524
• /
• 2023

In hashing-based image retrieval, the hash code of a manipulated image is different from the original image, making it difficult to search for the same image. This paper proposes and evaluates a self-supervised deephashing model that generates perceptual hash codes from feature information such as texture, shape, and color of images. The comparison models are autoencoder-based variational inference models, but the encoder is designed with a fully connected layer, convolutional neural network, and transformer modules. The proposed model is a variational inference model that includes a SimAM module of extracting geometric patterns and positional relationships within images. The SimAM module can learn latent vectors highlighting objects or local regions through an energy function using the activation values of neurons and surrounding neurons. The proposed method is a representation learning model that can generate low-dimensional latent vectors from high-dimensional input images, and the latent vectors are binarized into distinguishable hash code. From the experimental results on public datasets such as CIFAR-10, ImageNet, and NUS-WIDE, the proposed model is superior to the comparative model and analyzed to have equivalent performance to the supervised learning-based deephashing model. The proposed model can be used in application systems that require low-dimensional representation of images, such as image search or copyright image determination.

• Journal of Power Electronics
• /
• v.5 no.2
• /
• pp.99-103
• /
• 2005

In this paper, a novel circuit topology of a three-winding coupling inductor-assisting a high-frequency PWM boost chopper type DC-DC power converter with a high boost voltage conversion ratio and low switch voltage stress is proposed for the new energy interfaced DC power conditioner in solar photovoltaic and fuel cell generation systems. The operating principle in a steady state is described by using its equivalent circuits under the practical condition of energy processing of a lossless capacitive snubber. The newly-proposed power MOSFET boost chopper type DC-DC power converter with the three-winding coupled inductor type transformer and a single lossless capacitor snubber is built and tested for an output power of 500W. Utilizing the lower voltage and internal resistance power MOSFET switch in the proposed PWM boost chopper type DC-DC power converter can reduce the conduction losses of the active power switch compared to the conventional model. Therefore, the total actual power conversion efficiency under a condition of the nominal rated output power is estimated to be 81.1 ％, which is 3.7％ higher than the conventional PWM boost chopper DC power conversion circuit topology.

• Journal of Power Electronics
• /
• v.12 no.1
• /
• pp.145-156
• /
• 2012

These days, the application of electronic power transformers (EPTs) is expanding in place of ordinary power transformers. These transformers can transmit power via three or four wire converters. Their dynamic performance is extremely important, due to their complex structure. In this paper, a new method is proposed for improving the dynamic performance of distribution electronic power transformers (DEPT) by using sliding mode control (SMC). Hence, to express the dynamic characteristics of a system, different factors such as the voltage unbalance, voltage sag, voltage harmonics and voltage flicker in the system primary side are considered. The four controlling aims of the improvement in dynamic performance include: 1) maintaining the input currents so that they are in sinusoidal form and in phase with the input voltages so they have a unity power factor, 2) keeping the dc-link voltage within the reference amount, 3) keeping the output voltages at a fixed amount and 4) keeping the output voltages in sinusoidal and symmetrical forms. Simulation results indicate the potential and capability of the proposed method in improving DEPT behavior.

• Proceedings of the KIEE Conference
• /
• 1996.07a
• /
• pp.101-103
• /
• 1996

Multiply-twist cable is used for a large capacity superconducting cable because it is helpful to reduce AC losses and to increase transport current. In a multiply-twisted cable, the axis of a strand does not coincide with that of cable. Therefore, the longitudinal field is generated by the transport current. The longitudinal field changes the current distribution in the strand and generates additional AC loss. This paper calculates the longitudinal field that is applied to a strand in the multiply-twisted cable. Current distribution of a strand in the cable is also presented. 2nd level superconducting cable is chosen as an analysis model, whose current capacity is 2000A. Calculation result shows the longitudinal field cannot be neglected in low field machines such as superconducting transformer.

• The Proceeding of the Korean Institute of Electromagnetic Engineering and Science
• /
• v.3 no.1
• /
• pp.3-10
• /
• 1992

Flux-drive waveguide phase shifter is designed by twin-slab model. Measured differential phase shifts are smaller than the theoretical values by 8-9 percents. Measured insertion loss and VSWR of the phase shifter using TT73-2200 ferite are less than 0.45dB and 1.25 respectively, within pass band. The phase shifter using double-setup transformer shows wider bandwidth characteristics. Finally the reduced-height waveguide phase shifter using TT3-2900 ferrite shows very efficient suppression of higher-oreder modes.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.63 no.8
• /
• pp.1055-1063
• /
• 2014

Power losses of a 1-stage DC-DC converter and 2-stage DC-DC converter are compared in this paper. A phase-shift full-bridge DC-DC converter is considered as 1-stage topology. This topology has disadvantages in the stress of rectifier diodes because of the resonance between the leakage inductor of the transformer and the junction capacitor of the rectifier diode. 2-stage topology is composed of an LLC resonant full-bridge DC-DC converter and buck converter. The LLC resonant full-bridge DC-DC converter does not need an RC snubber circuit of the rectifier diode. However, there is the drawback that the switching loss of the buck converter is large due to the hard switching operation. To reduce the switching loss of the buck converter, SiC MOSFET is used. This paper analyzes and compares power losses of two topologies considering temperature condition. The validity of the power loss analysis and calculation is verified by a PSIM simulation model.

• The Transactions of the Korean Institute of Power Electronics
• /
• v.25 no.4
• /
• pp.286-292
• /
• 2020

This study proposes a power decoupled multi-port dual-active-bridge (DAB) DC-DC converter employing multiple transformers. Conventional multiport DAB DC-DC converters experience a power coupling issue from the use of a single transformer, which essentially requires complex power decoupling control. To solve this issue, a multiport DAB DC-DC converter employing multiple transformers is proposed to decouple output power without additional complex control algorithms. The proposed converter uses multiple transformers that can expand output ports easily. Therefore, transformers and the proposed multi-port DAB converter can be designed simply. In addition, the number of coupling inductors can be reduced in the proposed three-port DAB converter compared with that in conventional multiport DAB converters. The power decoupling characteristics and equivalent circuit of the proposed converter are analyzed using theoretical model approaches. Finally, a 3-kW laboratory prototype is developed to verify the effectiveness of the proposed converter.

• Wind and Structures
• /
• v.20 no.2
• /
• pp.315-326
• /
• 2015

To overcome the drawbacks of the traditional contact-type sensor for structural displacement measurement, the vision-based technology with the aid of the digital image processing algorithm has received increasing concerns from the community of structural health monitoring (SHM). The advanced vision-based system has been widely used to measure the structural displacement of civil engineering structures due to its overwhelming merits of non-contact, long-distance, and high-resolution. However, seldom currently-available vision-based systems are capable of realizing the synchronous structural displacement measurement for multiple points on the investigated structure. In this paper, the method for vision-based multi-point structural displacement measurement is presented. A series of moving loading experiments on a scale arch bridge model are carried out to validate the accuracy and reliability of the vision-based system for multi-point structural displacement measurement. The structural displacements of five points on the bridge deck are measured by the vision-based system and compared with those obtained by the linear variable differential transformer (LVDT). The comparative study demonstrates that the vision-based system is deemed to be an effective and reliable means for multi-point structural displacement measurement.

• The Transactions of the Korean Institute of Power Electronics
• /
• v.6 no.4
• /
• pp.317-324
• /
• 2001

This paper proposes a SSSC based on 3-level half-bridge inverters. The dynamic characteristic of the proposed SSSC was analyzed by EMTP simulation and a scaled hardware model, assuming that the SSSC is inserted in the transmission line of the one-machine-infinite-but power system. The proposed SSC has six 3-level half-bridge inverters per phase, which operates in PWM mode. The proposed SSSC generates a quasi-sinusoidal output voltage by 90 degree phase shift to the line current. The proposed SSSC does not require the coupling transformer for voltage injection, and has a flexibility in operation voltage by increasing the number of series connection.