• Electronics and Telecommunications Trends
• /
• v.39 no.4
• /
• pp.82-92
• /
• 2024

This paper introduces the technological development trends in on-device SLMs (Small Language Models). Large Language Models (LLMs) based on the transformer model have gained global attention with the emergence of ChatGPT, providing detailed and sophisticated responses across various knowledge domains, thereby increasing their impact across society. While major global tech companies are continuously announcing new LLMs or enhancing their capabilities, the development of SLMs, which are lightweight versions of LLMs, is intensely progressing. SLMs have the advantage of being able to run as on-device AI on smartphones or edge devices with limited memory and computing resources, enabling their application in various fields from a commercialization perspective. This paper examines the technical features for developing SLMs, lightweight technologies, semiconductor technology development trends for on-device AI, and potential applications across various industries.

• Proceedings of the KIPE Conference
• /
• 2001.10a
• /
• pp.295-300
• /
• 2001

A variety of high voltage DC power supplies employing the high frequency inverter are difficult to achieve soft switching considering a quick response and no overshoot response under the wide load variation ranges which are used in medical-use x-ray high voltage generator from 20kV to 150kV in the output voltage and from 0.5mA to 1250mA, respectively. The authors develops soft switching high voltage DC power supply designed for x-ray power generator applications, which uses series resonant inverter circuit topology with a multistage voltage multiplier instead of a conventional high voltage diode rectifier connected to the second-side of a high-voltage transformer with a large turn ratio. A constant on-time dual mode frequency control scheme operating under a principle of zero-current soft switching commutation is described. Introducing the multistage voltage multiplier, the secondary transformer turn-numbers and stray capacitance of high-voltage transformer is effective to be greatly reduced. It is proved that the proposed high-voltage converter topology with dual mode frequency modulation mode control scheme is able to be the transient response and steady-state performance in high-voltage x-ray tube load. The effectiveness of this high voltage converter is evaluated and discussed on the basis of simulation analysis and observed data in experiment.

• Analytical Science and Technology
• /
• v.18 no.3
• /
• pp.206-215
• /
• 2005

Polychlorinated biphenyls (PCBs) were commercially produced as complex mixtures beginning in 1929. The PCBs manufactured commercially are known by a variety of trade names including; Aroclor (USA), Phenoclor (France), Kaneclor (Japan), Sovol (USSR) and so on. PCBs are a class of 209 congeners that were widely used in a wide variety of applications, including dielectric fluid in transformers and large capacitors; heat transfer fluids; hydraulic fluids; lubricating and cutting oils; and as additives in pesticides, paints, adhesives, sealants, and plastics. The quantification methods of peak matching and coefficient comparison were compared using the Aroclor 1242, 1248, 1254, 1260 standards. Also, six transformer oils were analyzed as a main source of polychlorinated biphenyls (PCBs) emission into the environment. The transformer oils contained the pure and mixed of Aroclor 1242, Aroclor 1254, and Aroclor 1260. The analytical results using two quantification methods showed the little difference between the measured results.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.9 no.3
• /
• pp.638-644
• /
• 2008

It is required small size electronic ballast to fullfill the design philosophy of miniaturizing in the application of slim lamps. However, the traditional magnetic ballasts operated at 50-60Hz have been suffered from noticeable flicker, high loss, large crest factor and heavy weight. In this study, in order to solve these problems, It was proposed for driving 35W Class fluorescent lamp in the new type of electronic ballast, which is composed of rectifier, active power factor corrector, series resonant half bridge inverter and piezoelectric transformer. A 35W class(T5 class) fluorescent lamp is driven to successful by the fabricated ballast with piezoelectric transformer. Experimental results, It is proposed to driving the lamp using a electronic ballast at operating frequency of $75{\sim}79kHz$ approximatively. It is operated after for 25 min that were obtained good results of the input power factor of 0.95 and efficiency of 86%, respectively.

• Journal of Power Electronics
• /
• v.17 no.6
• /
• pp.1625-1636
• /
• 2017

A power electronic transformer (PET) based on the cascaded H-bridge (CHB) and the isolated bidirectional DC/DC converter (IBDC) is capable of accommodating a large scale battery energy storage system (BESS) in the medium-voltage grid, and is referred to as a power electronic transformer based battery energy storage system (PET-BESS). This paper investigates the PET-BESS and proposes a coordinative control strategy for it. In the proposed method, the CHB controls the power flow and the battery state-of-charge (SOC) balancing, while the IBDC maintains the dc-link voltages with feedforward implementation of the power reference and the switch status of the CHB. State-feedback and linear quadratic Riccati (LQR) methods have been adopted in the CHB to control the grid current, active power and reactive power. A hybrid PWM modulating method is utilized to achieve SOC balancing, where battery SOC sorting is involved. The feedforward path of the power reference and the CHB switch status substantially reduces the dc-link voltage fluctuations under dynamic power variations. The effectiveness of the proposed control has been verified both by simulation and experimental results. The performance of the PET-BESS under bidirectional power flow has been improved, and the battery SOC values have been adjusted to converge.

• Smart Media Journal
• /
• v.11 no.4
• /
• pp.30-37
• /
• 2022

Disease threatens plant growth and recognizing the type of disease is essential to making a remedy. In recent years, deep learning has witnessed a significant improvement for this task, however, a large volume of labeled images is one of the requirements to get decent performance. But annotated images are difficult and expensive to obtain in the agricultural field. Therefore, designing an efficient and effective strategy is one of the challenges in this area with few labeled data. Transfer learning, assuming taking knowledge from a source domain to a target domain, is borrowed to address this issue and observed comparable results. However, current transfer learning strategies can be regarded as a supervised method as it hypothesizes that there are many labeled images in a source domain. In contrast, unsupervised transfer learning, using only images in a source domain, gives more convenience as collecting images is much easier than annotating. In this paper, we leverage unsupervised transfer learning to perform plant disease recognition, by which we achieve a better performance than supervised transfer learning in many cases. Besides, a vision transformer with a bigger model capacity than convolution is utilized to have a better-pretrained feature space. With the vision transformer-based unsupervised transfer learning, we achieve better results than current works in two datasets. Especially, we obtain 97.3% accuracy with only 30 training images for each class in the Plant Village dataset. We hope that our work can encourage the community to pay attention to vision transformer-based unsupervised transfer learning in the agricultural field when with few labeled images.

• Journal of Korea Water Resources Association
• /
• v.57 no.7
• /
• pp.437-449
• /
• 2024

Rainfall-runoff prediction studies using deep learning while considering catchment attributes have been gaining attention. In this study, we selected two models: the Transformer model, which is suitable for large-scale data training through the self-attention mechanism, and the LSTM-based multi-state-vector sequence-to-sequence (LSTM-MSV-S2S) model with an encoder-decoder structure. These models were constructed to incorporate catchment attributes and predict the inflow of 10 multi-purpose dam watersheds in South Korea. The experimental design consisted of three training methods: Single-basin Training (ST), Pretraining (PT), and Pretraining-Finetuning (PT-FT). The input data for the models included 10 selected watershed attributes along with meteorological data. The inflow prediction performance was compared based on the training methods. The results showed that the Transformer model outperformed the LSTM-MSV-S2S model when using the PT and PT-FT methods, with the PT-FT method yielding the highest performance. The LSTM-MSV-S2S model showed better performance than the Transformer when using the ST method; however, it showed lower performance when using the PT and PT-FT methods. Additionally, the embedding layer activation vectors and raw catchment attributes were used to cluster watersheds and analyze whether the models learned the similarities between them. The Transformer model demonstrated improved performance among watersheds with similar activation vectors, proving that utilizing information from other pre-trained watersheds enhances the prediction performance. This study compared the suitable models and training methods for each multi-purpose dam and highlighted the necessity of constructing deep learning models using PT and PT-FT methods for domestic watersheds. Furthermore, the results confirmed that the Transformer model outperforms the LSTM-MSV-S2S model when applying PT and PT-FT methods.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.16 no.10
• /
• pp.2230-2236
• /
• 2012

Currently, the AC-to-DC power conversion system using diode rectifiers is mainly used in large vessels. Also, to reduce the total harmonic distortion(THD) of current and voltage, this system requires an additional phase-shifting transformer which can be powered multi-pulses. In this case, due to the installation of the transformer, the spatial or economic loss occurs. This paper presents a novel active rectification scheme using silicon controlled rectifier(SCR) or insulated gate bipolar transistor(IGBT) devices on behalf of the diode rectifiers which are currently operating in large vessels such as LNG Carrier(LNGC). The proposed system can use the low voltage source and reduce current and voltage harmonics generated by nonlinear loads connected to the power distribution bus and save economic costs by removing the phase-shifting transformers which are used in conventional system. Computer simulations are performed under the electric propulsion system which is operating in current large vessel. The results are shown in support of the improvement of THD included in the current and voltage wave forms of propulsion motor.

• Proceedings of the KSR Conference
• /
• 2005.05a
• /
• pp.794-799
• /
• 2005

In this paper, the inductive power collector using electromagnetic induction for the PRT(Personal Rapid Transit) system is suggested and some ideas for power collector design to improve the power transfer performance are presented. The proposed the inductive power collector is used for the PRT system, which has a large air-gap and demands a large electrical power capability. But, low output power is generated due to a loosely coupled characteristic of the large air-gap. Therefore, double layer construction of secondary winding. which was divided in half to increase both output current and output voltage was suggested. Also, model of power collector and parallel winding structure are presented, in addition, the performance of inductive power collector to alignment condition between the primary power line and the inductive power transformer was verified by computer simulation of 2kW model.

• Journal of Mechanical Science and Technology
• /
• v.19 no.8
• /
• pp.1576-1581
• /
• 2005

Generally, the optical components are fabricated by grinding, lapping, and polishing. And, those processes take long time to obtain such a high surface quality. Therefore, in the case of large optical component, the on-machine inspection (OMI) is essential. Because, the work piece is fragile and difficult to set up for fabricating and measuring. This paper is concerned about a swing-arm method for measuring surface profile of large optical concave mirror. The measuring accuracy and uncertainty for suggested method are studied. The experimental results show that this method is useful specially in lapping process with the accuracy of $3\~5\;{\mu}m$. Those inspection data are provided for correcting the residual figuring error in lapping or polishing processes.