• 제목/요약/키워드: Input Out Model

검색결과 777건 처리시간 0.028초

입력 변이에 따른 딥러닝 모델 취약점 연구 및 검증 (Analysis of Deep Learning Model Vulnerability According to Input Mutation)

  • 김재욱;박래현;권태경
    • 정보보호학회논문지
    • /
    • 제31권1호
    • /
    • pp.51-59
    • /
    • 2021
  • 딥러닝 모델은 변이를 통해 훈련 데이터에서 벗어난 입력으로부터 잘못된 예측 결과를 산출할 수 있으며 이는 자율주행, 보안 분야 등에서 치명적인 사고로 이어질 수 있다. 딥러닝 모델의 신뢰성 보장을 위해서는 다양한 변이를 통해 예외적인 상황에 대한 모델의 처리 능력이 검증되어야 한다. 하지만, 기존 연구가 제한된 모델을 대상으로만 수행되었으며, 여러 입력 변이 유형에 구분을 짓지 않고 사용했다. 본 연구에서는 딥러닝 검증 데이터 세트로 널리 사용되고 있는 CIFAR10 데이터 세트를 기반으로 다양한 상용화된 모델과 추가 버전을 포함하여 총 6개의 모델에 대한 신뢰성 검증을 수행한다. 이를 위해 실생활에서 발생할 수 있는 6가지 유형의 입력 변이 알고리즘을 다양한 파라미터와 함께 데이터 세트에 개별적으로 적용하여 각각에 대한 모델의 정확도를 비교함으로써 특정 변이 유형과 관련된 모델의 취약점을 구체적으로 파악한다.

Design of Subband Codecs Using Optimized Vector Quantizer

  • Jee, Innho
    • The Journal of the Acoustical Society of Korea
    • /
    • 제15권2E호
    • /
    • pp.33-38
    • /
    • 1996
  • This paper provides an approach for representing an optimum vector quantizer by a scalar nonlinear gain-plus-additive noise model. The validity and accuracy of this analytic model is confirmed by comparing the calcuated model quantization errors with actual simulation of the optimum Linde-Buzo-Gray(LBG) vector quantizer. Using this model we frm MSE measure of an M-band filter bank codec in terms of the equivalent scalar quantizatin model and find the optimum FIR filter coefficients for each channel in the M-band structure for a given bit rate, given filter length, and given input signal correlation model. Specific design examples are worked out for 4-tap filters in the two-band paraunitary case. These theoretical results are confirmed by extensive Monte Carlo simulation.

  • PDF

유전알고리즘에 의한 퍼지모델기반의 상태관측기 설계 (Design of a Fuzzy Model-Based State Observer Using GAs)

  • 이현식;손영득;김종화;유영호;하윤수;진강규
    • Journal of Advanced Marine Engineering and Technology
    • /
    • 제25권1호
    • /
    • pp.162-170
    • /
    • 2001
  • This paper presents a scheme for designing a fuzzy model-bsaed state observer for nonlinear system. For this scheme, a Tagaki-Sugeno type fuzzy model whose consequent part is of the state space form is obtained. In describes the locally linear input/output relationship of a system. The parameters of the fuzzy model are adjusted using a genetic algorithm. Then. fuzzy full-order and reduced-order state observers are designed based on the fuzzy model. A set of simulation works is carried out to demonstrate the effectiveness of the proposed scheme.

  • PDF

상수관망 모형의 비교 분석 연구 (A Comparison Study on Water Network Models)

  • 김준현;나탈리아 야꾸니나
    • 환경영향평가
    • /
    • 제19권3호
    • /
    • pp.307-314
    • /
    • 2010
  • Brebbia's model has been analyzed to develop the appropriate waterworks management system in Korea, and compared with the conventional models such as EPANET, WaterCad, and InfoWorks. The hydraulic theory of the models was analyzed. Each model's numerical techniques, required parameters, input data and operational methodologies, restrictions, practical applicability and other aspects were investigated. In order to check the validity of Brebbia model, the comparative analysis with EPANET, WaterCAD, and InfoWorks models was performed for linear and nonlinear cases. To find out advantages and disadvantages of each model, the modeling was performed for a simple network and for more complicated A city waterworks system, and the three models applicability was examined. Finally, optimal modeling technique and a model suitable for the use in Korea was suggested, and the problems related to present projects of waterworks management system in Korea were analyzed.

실시간 다물체 차량 동역학 모델 개발 및 구현 (Development and Implementation of Real Time Multibody Vehicle Dynamics Model)

  • 오영석;김성수
    • 대한기계학회논문집A
    • /
    • 제25권5호
    • /
    • pp.834-840
    • /
    • 2001
  • A real time multibody vehicle dynamics model has been developed and implemented using a subsystem synthesis method based on recursive formulation. To verify real time simulation capability the developed model has been applied to HMMWV(High Mobility Multipurpose Wheeled Vehicle) with steering system. For the kinematically driven steering system, the coupled front suspension-steering subsystem can be decoupled into two SLA suspension subsystems, which improves the efficiency of simulation. To investigate theoretical efficiency, operational counting method has been also employed to compare the proposed model with the conventional recursive dynamics model. Various simulations such as unsymmetric bump run, step steering(J-turn) and sine steering input test have been carried out to verify the real time feasibility of the proposed model.

선박화재의 인명안전평가 해석 (A Study on Evacuee′s Risk Assessment under Ship′s Fire)

  • 양영순;정정호;이재옥;공수철;여인철
    • 한국전산구조공학회:학술대회논문집
    • /
    • 한국전산구조공학회 2001년도 봄 학술발표회 논문집
    • /
    • pp.241-247
    • /
    • 2001
  • A Fire Safety Evaluation Module(FSEM), which quantitatively evaluates the risk of evacuees when fire occurs in buildings or ships, is presented in this paper. The developed FSEM can be applied to multi-room structure. Basic input data for the FSEM are prepared by fire model and evacuation model. CFAST which is one of the existing fire models is used as fire model and MonteDEM evacuation model was developed for evacuation model, respectively. MonteDEM evacuation model makes use of distinct element method and Monte-Carlo simulation, and it can also take into consideration ground inclination by ship motions in order to simulate the real situation of evacuation. Some typical situations are modelled for illustrative examples and quantitative assessment of evacuee's risk under fire accident is carried out.

  • PDF

더미변수(Dummy Variable)를 포함하는 다변수 시계열 모델을 이용한 단기부하예측 (Short-Term Load Forecasting Using Multiple Time-Series Model Including Dummy Variables)

  • 이경훈;김진오
    • 대한전기학회논문지:전력기술부문A
    • /
    • 제52권8호
    • /
    • pp.450-456
    • /
    • 2003
  • This paper proposes a multiple time-series model with dummy variables for one-hour ahead load forecasting. We used 11 dummy variables that were classified by day characteristics such as day of the week, holiday, and special holiday. Also, model specification and selection of input variables including dummy variables were made by test statistics such as AIC(Akaike Information Criterion) and t-test statistics of each coefficient. OLS (Ordinary Least Squares) method was used for estimation and forecasting. We found out that model specifications for each hour are not identical usually at 30% of optimal significance level, and dummy variables reduce the forecasting error if they are classified properly. The proposed model has much more accurate estimates in forecasting with less MAPE (Mean Absolute Percentage Error).

An Efficient and High-gain Inverter Based on The 3S Inverter Employs Model Predictive Control for PV Applications

  • Abdel-Rahim, Omar;Funato, Hirohito;Junnosuke, Haruna
    • Journal of Electrical Engineering and Technology
    • /
    • 제12권4호
    • /
    • pp.1484-1494
    • /
    • 2017
  • We present a two-stage inverter with high step-up conversion ratio engaging modified finite-set Model Predictive Control (MPC) for utility-integrated photovoltaic (PV) applications. The anticipated arrangement is fit for low power PV uses, the calculated efficiency at 150 W input power and 19 times boosting ratio was around 94%. The suggested high-gain dc-dc converter based on Cockcroft-Walton multiplier constitutes the first-stage of the offered structure, due to its high step-up ability. It can boost the input voltage up to 20 times. The 3S current-source inverter constitutes the second-stage. The 3S current-source inverter hires three semiconductor switches, in which one is functioning at high-frequency and the others are operating at fundamental-frequency. The high-switching pulses are varied in the procedure of unidirectional sine-wave to engender a current coordinated with the utility-voltage. The unidirectional current is shaped into alternating current by the synchronized push-pull configuration. The MPC process are intended to control the scheme and achieve the subsequent tasks, take out the Maximum Power (MP) from the PV, step-up the PV voltage, and introduces low current with low Total Harmonic Distortion (THD) and with unity power factor with the grid voltage.

SCM환경에서의 물류센터의 최적 서비스 수준 결정 방법 (A study on the Method to Determine Optimal Service Level of a Distribution Center in Supply Chain Management Environment)

  • 조용욱;박명규
    • 대한안전경영과학회지
    • /
    • 제3권3호
    • /
    • pp.55-64
    • /
    • 2001
  • The main objective of this research is to develop a model to select the optimal input service level for a distribution center-multi branch inventory distribution system. With the continuous review policy, the distribution center places an order for specific order quantity to an outside supplier, and the order quantity is replenished after a certain lead time Also, each branch places an order for particular order quantity to the distribution center to satisfy the customer demands, and receives the replenishment after a lead time. When an out of stock condition occurs during an order cycle, a backorder is placed to the upper level to fill the unfilled demands. With these situation, variable demand and variable lead time are used for better industrial practice. Further, actual lead times with a generic lead time distribution are used in developing the control model. Under the actual lead time model, the customer service measures actually attained for the distribution center and each branch are explained as the effective customer service measures. Thus, throughout the optimal control (using computer search procedures), we can select the optimal input service levels for the distribution center and each branch to attain the effective service levels for each branch which is consistent with the goal level of service for each branch. At the same time, the entire distribution system keeps minimum inventories.

  • PDF

박판 점 곡직 시 변형 특성에 관한 연구 (Study on the Deformation Behavior by Spot Heating for thin plate)

  • 장경복;박중구;조시훈;장태원
    • 대한용접접합학회:학술대회논문집
    • /
    • 대한용접접합학회 2005년도 춘계학술발표대회 개요집
    • /
    • pp.285-287
    • /
    • 2005
  • During fabrication of deck house block in passenger ships, the problem of unexpected large deformation and distortion frequently occurs. In this case, line and spot heating method were widely used to correct the distortion of thin plate structure. Spot heating was especially used for the case under 5mm thickness. Few papers are available on the working conditions of spot heating method but only little information on deformation control. In this study, evaluation was carried out on the temperature distribution of spot heating methods using FEA and practical experiments for various heating time. IIn FEA, heat input model was established using Tsuji's double Gaussian heat input mode (Tsuji, I., 1988). This model was verified by comparing with experimental data. Also radial shrinkage and angular distortion due to spot heating were determined and compared with experimental results. Thermo elasto-plastic analysis was performed using commercial FE code, MSC/MARC. Radial shrinkage and angular distortion were measured using 3D measuring apparatus. Based on these results, simplified analysis model for deformation by spot heating was established.

  • PDF