• 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.

• ETRI Journal
• /
• v.36 no.2
• /
• pp.293-300
• /
• 2014

As the system-on-chip (SoC) design becomes more complex, the test costs are increasing. One of the main obstacles of a test cost reduction is the limited number of test channels of the ATE while the number of pins in the design increases. To overcome this problem, a new test architecture using a channel sharing compliant with IEEE Standard 1149.1 and 1500 is proposed. It can significantly reduce the pin count for testing a SoC design. The test input data is transmitted using a test access mechanism composed of only input pins. A single test data output pin is used to measure the sink values. The experimental results show that the proposed architecture not only increases the number of sites to be tested simultaneously, but also reduces the test time. In addition, the yield loss owing to the proven contact problems can be reduced. Using the new architecture, it is possible to achieve a large test time and cost reduction for complex SoC designs with negligible design and test overheads.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.24 no.2 s.173
• /
• pp.506-517
• /
• 2000

One dimensional performance model has been used for the design of torque converter. The model is based on the concept of constant mean flow path and constant flow angle. These constant-assumed para meters make the design procedure to be simple. In practice, some parameters are usually replaced with geometric raw data and, the constant experiential correction factors have been used to minimize the design error. These factors have no definite physical meaning and so they cannot be applied confidently to the other design condition. In this study, the detail dynamic model of torque converter is presented to establish the theoretical background of correction factors. To verify the validity of theoretical model, steady state performance test was carried out on the several input speed. The oil temperature effect on the performance is analysed and adjusted. The constant equivalent flow angles are determined at a part of performance region by comparing the theoretical model and the test data. The sensitivity of correction factors to the input speeds are studied and the change of torus flow is presented.

• Journal of the Korean Society for Precision Engineering
• /
• v.16 no.6
• /
• pp.123-132
• /
• 1999

This paper describes a research work of developing computer-aided design of lead frame, semiconductor, with blanking operation which is very precise for progressive working. Approach to the system is based on the knowledge-based rules. Knowledge for the system is formulated from pasticity theories, experimental results and the empirical knowledge of field experts. This system has been written in AutoLISP on the AutoCAD using a personal computer and in I-DEAS Drafting Programming Language on the I-DEAS Master Series Drafting with Workstation, HP9000/715(64). Transference of data between AutoCAD and I-DEAS Master Series Drafting is accomplished by DXF(drawing exchange format) and IGES(initial graphics exchange specification) methods. This system is composed of five modules, which are input and shape treatment, production feasibility check, strip-layout, data-conversion and die-layout modules. The process planning and Die design system is designed by considering several factors, such as complexities of blank geometry, punch profiles, and the availability of a press equipment and standard parts. This system provides its efficiecy for strip-layout, and die design for lead frame, semiconductor.

• Journal of the Korean Society for Precision Engineering
• /
• v.13 no.5
• /
• pp.95-103
• /
• 1996

This paper describes a computer aided design system for spur and helical gears. To establish the appropriate program, an integrate approach based on a rule-base system was adopted. This system is implemented on the personal computer and its environment is a commercial CAD package called AutoCAD. This system includes a main program and five sub-modules such as data input module, tooth profile drawing module, strength calculation module, and drawing edit module. In the main program, all the sub-modules are loaded and the type of gear and tooth profile are selected. In the data input module, the variables which are necessary to the design of gear are selected from the database. In the drawing module, from the calculated results, the required gear tooth is produced on the screen. The developed system that aids gear designer provides powerful capabilities for gear design.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2003.10a
• /
• pp.125-130
• /
• 2003

This Paper describes a Nonlinear adoptive noise canceller using Neural Network for Machine Tools Controller System. Back-Propagation Learning Algorithm based MLP (Multi Layer Perceptron)is used an adaptive filters. In this Paper. it assume that the noise of primary input in the adaptive noise canceller is not the same characteristic as that of the reference input. Experimental results show that the neural network base noise canceller outperforms the linear noise canceller. Especially to make noise cancel close to realtime, Primary Input is divided by Unit and each divided pan is processed for very short time than all the processed data are unified to whole data.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2004.10a
• /
• pp.345-351
• /
• 2004

This Paper describes a nonlinear adaptive noise filter using neural network for digital controller system. Back-Propagation Learning Algorithm based MLP (Multi Layer Perceptron)is used an adaptive filters. In this paper. it assume that the noise of primary input in the adaptive noise canceller is not the same characteristic as that of the reference input. Experimental reaults show that the neural network base noise canceller outperforms the linear noise canceller. Especially to make noise cancel close to realtime, Primary input is divided by unit and each divided part is processed for very short time than all the processed data are unified to whole data.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 1999.10a
• /
• pp.378-386
• /
• 1999

The integrated system for the structural analysis of space frame is made up 4 modules ; pre-process module, structural analysis module, optimum member design module and post-process module. Re-process module of these 4 modules involves data input module and structure modeling module. This study is to develope an efficient modeling program as a basic for development of pre-process module. This modeling program generates geometric information of space frame and performs the input fie form for structure analysis only by input general data. User can mode1 space frame efficiently within shut time by using this program.

• International Journal of CAD/CAM
• /
• v.7 no.1
• /
• pp.103-112
• /
• 2007

This paper presents a practical approach for surface reconstruction from unoriented point clouds. Instead of estimating local surface orientation, we first generate a set of depth images from the input point clouds, and a coarse mesh is then generated based on them by space carving techniques. The resultant mesh is progressively refined by local mesh refinement and optimization according to surface distance measure. A manifold mesh approximating the input points within an given tolerance is finally obtained. Our approach is easy to implement, but has the ability to outputs high quality meshes in different resolutions. We show that the proposed approach is not sensitive to several types of data disfigurement and is able to reconstruct models robustly from variance input data.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.23 no.6
• /
• pp.944-949
• /
• 1986

By improving the ability of arithmatic processing with an arithmatic processor in a microcomputer and realizing the data input system for real time analysis, an FFT analyzer that is usable within the range of audio frequency is designed and constructed. The input signal passes through a gain programmable pre-amplifier and anti-aliasing lowpass filter into an analogditital converter to be converted into digital form. The converted input data is processed by an Apple II microcomputer. The results of the processing are displayed using a microcomputer display unit and can be copied on a printer or stored in a floppy disk.