• 정보처리학회 논문지
• /
• 제13권2호
• /
• pp.76-90
• /
• 2024

인공지능 기술은 스마트 시티, 자율 주행, 의료 분야 등 다양한 분야에서 활용 가능성을 높이 평가받고 있으나, 정보주체의 개인정보 및 민감정보의 노출 문제로 모델 활용이 제한되고 있다. 이에 따라 데이터를 중앙 서버에 모아서 학습하지 않고, 보유 데이터셋을 바탕으로 일차적으로 학습을 진행한 후 글로벌 모델을 최종적으로 학습하는 분산 기계 학습의 개념이 등장하였다. 그러나, 분산 기계 학습은 여전히 협력하여 학습을 진행하는 과정에서 데이터 프라이버시 위협이 발생한다. 본 연구는 분산 기계 학습 연구 분야에서 프라이버시를 보호하기 위한 연구를 서버의 존재 유무, 학습 데이터셋의 분포 환경, 참여자의 성능 차이 등 현재까지 제안된 분류 기준들을 바탕으로 유기적으로 분석하여 최신 연구 동향을 파악한다. 특히, 대표적인 분산 기계 학습 기법인 수평적 연합학습, 수직적 연합학습, 스웜 학습에 집중하여 활용된 프라이버시 보호 기법을 살펴본 후 향후 진행되어야 할 연구 방향을 모색한다.

• 인터넷정보학회논문지
• /
• 제24권6호
• /
• pp.119-125
• /
• 2023

우크라이나-러시아 전을 통해 드론의 군사적 가치는 재평가되고 있으며, 북한은 '22년 말 대남 드론 도발을 통해 실제 검증까지 완료한 바 있다. 또한, 북한은 인공지능(AI) 기술의 드론 적용을 추진하고 있는 것으로 드러나 드론의 위협은 나날이 커지고 있다. 이에 우리 군은 드론작전사령부를 창설하고 다양한 드론 대응 체계를 도입하는 등 대 드론 체계 구축을 도모하고 있지만, 전력증강 노력이 타격체계 위주로 편중되어 군집드론 공격에 대한 효과적 대응이 우려된다. 특히, 도심에 인접한 공군 비행단은 민간 피해가 우려되어 재래식 방공무기의 사용 역시 극도로 제한되는 실정이다. 이에 본 연구에서는 AI기술이 적용된 적 군집드론의 위협으로부터 아 항공기의 생존성 향상을 위해 AI모델의 객체탐지 능력을 저해하는 소극방공 기법을 제안한다. 대표적인 적대적 머신러닝(Adversarial machine learning) 기술 중 하나인 적대적 예제(Adversarial example)를 레이저를 활용하여 항공기에 조사함으로써, 적 드론에 탑재된 객체인식 AI의 인식률 저하를 도모한다. 합성 이미지와 정밀 축소모형을 활용한 실험을 수행한 결과, 제안기법 적용 전 약 95%의 인식률을 보이는 객체인식 AI의 인식률을 제안기법 적용 후 0~15% 내외로 저하시키는 것을 확인하여 제안기법의 실효성을 검증하였다.

• 한국항행학회논문지
• /
• 제21권6호
• /
• pp.521-528
• /
• 2017

다수의 UAV가 다양한 임무를 수행하면서도 편대를 유지할 수 있도록 하는 집단적 지능을 갖춘 시스템을 구축하기 위해서는 AI에 관한 연구가 필수적이다. AI의 전형적인 접근 방법에는 전문가시스템을 비롯한 규칙기반의 논리 추론방식인 '하향식' 접근 방법과 인공신경회로망, Flocking Algorithm과 같이 단순 개체간의 부분적 상호작용을 통해 전체적인 행동이 결정되는 '상향식' 접근 방법이 있다. 기존의 Flocking Algorithm과 같은 연구에서는 개개인은 개별적인 임무를 수행 할 수 없다. 또한 UAV의 편대비행과 같은 연구에서는 편대의 부분적인 결함으로 발생하는 문제에 대해 유연하게 대처 할 수 없다. 본 논문에서는 다중에이전트 시스템을 통해 하향식 접근 방법과 상향식 접근 방법 간의 유기적 통합을 제시하고, 이를 통해, 유연한 임무수행이 가능한 편대 비행 알고리즘을 제시하였으며, 시뮬레이션을 통해 대형형성 및 충돌회피 등 유효성을 확인하였다.

• Computers and Concrete
• /
• 제32권2호
• /
• pp.149-163
• /
• 2023

One of the main design parameters traditionally utilized in projects of geotechnical engineering is the uniaxial compressive strength. The present paper employed three artificial intelligence methods, i.e., the stochastic fractal search (SFS), the multi-verse optimization (MVO), and the vortex search algorithm (VSA), in order to determine the compressive strength of concrete (CSC). For the same reason, 1030 concrete specimens were subjected to compressive strength tests. According to the obtained laboratory results, the fly ash, cement, water, slag, coarse aggregates, fine aggregates, and SP were subjected to tests as the input parameters of the model in order to decide the optimum input configuration for the estimation of the compressive strength. The performance was evaluated by employing three criteria, i.e., the root mean square error (RMSE), mean absolute error (MAE), and the determination coefficient (R²). The evaluation of the error criteria and the determination coefficient obtained from the above three techniques indicates that the SFS-MLP technique outperformed the MVO-MLP and VSA-MLP methods. The developed artificial neural network models exhibit higher amounts of errors and lower correlation coefficients in comparison with other models. Nonetheless, the use of the stochastic fractal search algorithm has resulted in considerable enhancement in precision and accuracy of the evaluations conducted through the artificial neural network and has enhanced its performance. According to the results, the utilized SFS-MLP technique showed a better performance in the estimation of the compressive strength of concrete (R²=0.99932 and 0.99942, and RMSE=0.32611 and 0.24922). The novelty of our study is the use of a large dataset composed of 1030 entries and optimization of the learning scheme of the neural prediction model via a data distribution of a 20:80 testing-to-training ratio.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권2호
• /
• pp.564-582
• /
• 2015

This paper proposes a novel swarm intelligence optimization method which integrates bacterial foraging optimization (BFO) with quantum computing, called quantum bacterial foraging optimization (QBFO) algorithm. In QBFO, a multi-qubit which can represent a linear superposition of states in search space probabilistically is used to represent a bacterium, so that the quantum bacteria representation has a better characteristic of population diversity. A quantum rotation gate is designed to simulate the chemotactic step for the sake of driving the bacteria toward better solutions. Several tests are conducted based on benchmark functions including multi-peak function to evaluate optimization performance of the proposed algorithm. Numerical results show that the proposed QBFO has more powerful properties in terms of convergence rate, stability and the ability of searching for the global optimal solution than the original BFO and quantum genetic algorithm. Furthermore, we examine the employment of our proposed QBFO for cognitive radio spectrum allocation. The results indicate that the proposed QBFO based spectrum allocation scheme achieves high efficiency of spectrum usage and improves the transmission performance of secondary users, as compared to color sensitive graph coloring algorithm and quantum genetic algorithm.

• Journal of Electrical Engineering and Technology
• /
• 제6권2호
• /
• pp.174-181
• /
• 2011

This paper proposes a novel binary ant colony optimization (NBACO) method. The proposed NBACO is based on the concept and principles of ant colony optimization (ACO), and developed to solve the binary and combinatorial optimization problems. The concept of conventional ACO is similar to Heuristic Dynamic Programming. Thereby ACO has the merit that it can consider all possible solution sets, but also has the demerit that it may need a big memory space and a long execution time to solve a large problem. To reduce this demerit, the NBACO adopts the state probability matrix and the pheromone intensity matrix. And the NBACO presents new updating rule for local and global search. The proposed NBACO is applied to test power systems of up to 100-unit along with 24-hour load demands.

• Journal of Computing Science and Engineering
• /
• 제8권1호
• /
• pp.1-10
• /
• 2014

The Bacterial Foraging Optimization Algorithm is a swarm intelligence optimization algorithm. This paper first analyzes the chemotaxis, as well as elimination and dispersal operation, based on the basic Bacterial Foraging Optimization Algorithm. The elimination and dispersal operation makes a bacterium which has found or nearly found an optimal position escape away from that position, which greatly affects the convergence speed of the algorithm. In order to avoid this escape, the sphere of action of the elimination and dispersal operation can be altered in accordance with the generations of evolution. Secondly, we put forward an algorithm of an adaptive adjustment of step length we called improved bacterial foraging optimization (IBFO) after making a detailed analysis of the impacts of the step length on the efficiency and accuracy of the algorithm, based on chemotaxis operation. The classic test functions show that the convergence speed and accuracy of the IBFO algorithm is much better than the original algorithm.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권9호
• /
• pp.3496-3514
• /
• 2015

One of the challenging tasks in Mobile Ad hoc Network is to discover precise optimal routing solution due to the infrastructure-less dynamic behavior of wireless mobile nodes. Ant Colony Optimization, a swarm Intelligence technique, inspired by the foraging behaviour of ants in colonies was used in the past research works to compute the optimal path. In this paper, we propose a Recurrent Ant Colony Optimization (RECACO) that executes the actual Ant Colony Optimization iteratively based on recurrent value in order to obtain an optimal path convergence. Each iteration involves three steps: Pheromone tracking, Pheromone renewal and Node selection based on the residual energy in the mobile nodes. The novelty of our approach is the inclusion of new pheromone updating strategy in both online step-by-step pheromone renewal mode and online delayed pheromone renewal mode with the use of newly proposed metric named ELD (Energy Load Delay) based on energy, Load balancing and end-to-end delay metrics to measure the performance. RECACO is implemented using network simulator NS2.34. The implementation results show that the proposed algorithm outperforms the existing algorithms like AODV, ACO, LBE-ARAMA in terms of Energy, Delay, Packet Delivery Ratio and Network life time.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제6권9호
• /
• pp.2037-2051
• /
• 2012

Cognitive radio (CR) is envisioned as a promising paradigm of exploiting intelligence for enhancing efficiency of underutilized spectrum bands. In CR, the main concern is to reliably sense the presence of primary users (PUs) to attain protection against harmful interference caused by potential spectrum access of secondary users (SUs). In this paper, evolutionary algorithms, namely, particle swarm optimization (PSO) and genetic algorithm (GA) are proposed to minimize the total sensing decision error at the common soft data fusion (SDF) centre of a structurally-centralized cognitive radio network (CRN). Using these techniques, evolutionary operations are invoked to optimize the weighting coefficients applied on the sensing measurement components received from multiple cooperative SUs. The proposed methods are compared with each other as well as with other conventional deterministic algorithms such as maximal ratio combining (MRC) and equal gain combining (EGC). Computer simulations confirm the superiority of the PSO-based scheme over the GA-based and other conventional MRC and EGC schemes in terms of detection performance. In addition, the PSO-based scheme also shows promising convergence performance as compared to the GA-based scheme. This makes PSO an adequate solution to meet real-time requirements.

• Structural Engineering and Mechanics
• /
• 제56권3호
• /
• pp.385-407
• /
• 2015

This paper deals with the problem of the global stabilization for a class of tension leg platform (TLP) nonlinear control systems. It is well known that, in general, the global asymptotic stability of the TLP subsystems does not imply the global asymptotic stability of the composite closed-loop system. Finding system parameters for stabilizing the control system is also an issue need to be concerned. In this paper, we give additional sufficient conditions for the global stabilization of a TLP nonlinear system. In particular, we consider a class of NN based Takagi-Sugeno (TS) fuzzy TLP systems. Using the so-called parallel distributed compensation (PDC) controller, we prove that this class of systems can be globally asymptotically stable. The proper design of system parameters are found by a swarm intelligence algorithm called Evolved Bat Algorithm (EBA). An illustrative example is given to show the applicability of the main result.