• 정보보호학회논문지
• /
• 제33권6호
• /
• pp.881-891
• /
• 2023

최근에는 인공지능을 활용하여 악성 URL을 탐지하는 다양한 연구가 진행되고 있으며, 대부분의 연구 결과에서 높은 탐지 성능을 보였다. 그러나 고전 머신러닝을 활용하는 경우 feature를 분석하고 선별해야 하는 추가 비용이 발생하며, 데이터 분석가의 역량에 따라 탐지 성능이 결정되는 이슈가 있다. 본 논문에서는 이러한 이슈를 해결하기 위해 URL lexical feature를 자동으로 추출하는 딥러닝 모델의 일부가 고전 머신러닝 모델에 결합된 형태인 DL-ML Fusion Hybrid 모델을 제안한다. 제안한 모델로 직접 수집한 총 6만 개의 악성과 정상 URL을 학습한 결과 탐지 성능이 최대 23.98%p 향상되었을 뿐만 아니라, 자동화된 feature engineering을 통해 효율적인 기계학습이 가능하였다.

• 제어로봇시스템학회논문지
• /
• 제12권7호
• /
• pp.666-670
• /
• 2006

This paper presents the novel feature selection method for Emotion Recognition, which may include a lot of original features. Specially, the emotion recognition in this paper treated speech signal with emotion. The feature selection has some benefits on the pattern recognition performance and 'the curse of dimension'. Thus, We implemented a simulator called 'IFS' and those result was applied to a emotion recognition system(ERS), which was also implemented for this research. Our novel feature selection method was basically affected by Reinforcement Learning and since it needs responses from human user, it is called 'Interactive feature Selection'. From performing the IFS, we could get 3 best features and applied to ERS. Comparing those results with randomly selected feature set, The 3 best features were better than the randomly selected feature set.

• 한국멀티미디어학회논문지
• /
• 제22권2호
• /
• pp.203-209
• /
• 2019

The performance of an intelligent classifier for detecting malwares added to multimedia contents based on machine learning is highly dependent on the properties of feature set. Especially, in order to determine the malicious code in real time the size of feature set should be as short as possible without reducing the accuracy. In this paper, we introduce an optimal feature selection method to satisfy both high detection rate and the minimum length of feature set against the feature set provided by PEFeatureExtractor well known as a feature extraction tool. For the evaluation of the proposed method, we perform the experiments using Windows Portable Executables 32bits.

• 전기전자학회논문지
• /
• 제25권2호
• /
• pp.273-279
• /
• 2021

본 논문에서는 딥러닝 구조를 분할을 이용한 개별 학습을 수행하여 학습시간을 단축하는 알고리즘을 제안한다. 제안하는 알고리즘은 망 분류 기점 설정 과정, 특징 벡터 추출 과정, 특징 노이즈 제거 과정, 클래스 분류 과정 등의 4가지 과정으로 구성된다. 첫 번째로 망 분류 기점 설정 과정에서는 효과적인 특징 벡터 추출을 위한 망 구조의 분할 기점을 설정한다. 두 번째로 특징 벡터 추출 과정에서는 기존에 학습한 가중치를 사용하여 추가 학습 없이 특징 벡터를 추출한다. 세 번째로 특징 노이즈 제거 과정에서는 추출된 특징 벡터를 입력받아 각 클래스의 출력값을 학습하여 데이터의 노이즈를 제거한다. 네 번째로 클래스 분류 과정에서는 노이즈가 제거된 특징 벡터를 입력받아 다층 퍼셉트론 구조에 입력하고 이를 출력하고 학습한다. 제안된 알고리즘의 성능을 평가하기 위하여 Extended Yale B 얼굴 데이터베이스를 사용하여 실험 하였다. 실험 결과, 1회 학습에 소요되는 시간의 경우 제안하는 알고리즘이 기존 알고리즘 기준 40.7% 단축하였다. 또한 목표 인식률까지 학습 횟수가 기존 알고리즘과 비교하여 단축하였다. 실험결과를 통해 1회 학습시간과 전체 학습시간을 감소시켜 기존의 알고리즘보다 향상됨을 확인하였다.

• 대한전기학회논문지:시스템및제어부문D
• /
• 제52권5호
• /
• pp.277-282
• /
• 2003

In this paper we propose Using Higher Order Neuron on the Supervised Learning Machine of the Kohonen Feature Map. The architecture of proposed model adopts the higher order neuron in the input layer of Kohonen Feature Map as a Supervised Learning Machine. It is able to estimate boundary on input pattern space because or the higher order neuron. However, it suffers from a problem that the number of neuron weight increases because of the higher order neuron in the input layer. In this time, we solved this problem by placing the second order neuron among the higher order neuron. The feature of the higher order neuron can be mapped similar inputs on the Kohonen Feature Map. It also is the network with topological mapping. We have simulated the proposed model in respect of the recognition rate by XOR problem, discrimination of 20 alphabet patterns, Mirror Symmetry problem, and numerical letters Pattern Problem.

• Journal of Information Processing Systems
• /
• 제17권2호
• /
• pp.385-398
• /
• 2021

The complexity of deep learning models affects the real-time performance of gesture recognition, thereby limiting the application of gesture recognition algorithms in actual scenarios. Hence, a residual learning neural network based on a deep convolutional neural network is proposed. First, small convolution kernels are used to extract the local details of gesture images. Subsequently, a shallow residual structure is built to share weights, thereby avoiding gradient disappearance or gradient explosion as the network layer deepens; consequently, the difficulty of model optimisation is simplified. Additional convolutional neural networks are used to accelerate the refinement of deep abstract features based on the spatial importance of the gesture feature distribution. Finally, a fully connected cascade softmax classifier is used to complete the gesture recognition. Compared with the dense connection multiplexing feature information network, the proposed algorithm is optimised in feature multiplexing to avoid performance fluctuations caused by feature redundancy. Experimental results from the ISOGD gesture dataset and Gesture dataset prove that the proposed algorithm affords a fast convergence speed and high accuracy.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제18권6호
• /
• pp.1540-1561
• /
• 2024

This research proposes a novel approach for Tamil Handwritten Character Recognition (THCR) that combines feature selection and ensemble learning techniques. The Tamil script is complex and highly variable, requiring a robust and accurate recognition system. Feature selection is used to reduce dimensionality while preserving discriminative features, improving classification performance and reducing computational complexity. Several feature selection methods are compared, and individual classifiers (support vector machines, neural networks, and decision trees) are evaluated through extensive experiments. Ensemble learning techniques such as bagging, and boosting are employed to leverage the strengths of multiple classifiers and enhance recognition accuracy. The proposed approach is evaluated on the HP Labs Dataset, achieving an impressive 95.56% accuracy using an ensemble learning framework based on support vector machines. The dataset consists of 82,928 samples with 247 distinct classes, contributed by 500 participants from Tamil Nadu. It includes 40,000 characters with 500 user variations. The results surpass or rival existing methods, demonstrating the effectiveness of the approach. The research also offers insights for developing advanced recognition systems for other complex scripts. Future investigations could explore the integration of deep learning techniques and the extension of the proposed approach to other Indic scripts and languages, advancing the field of handwritten character recognition.

• 반도체디스플레이기술학회지
• /
• 제19권2호
• /
• pp.60-67
• /
• 2020

Recently, artificial intelligence related technologies including machine learning are being applied to various fields, and the demand is also increasing. In particular, with the development of AR, VR, and MR technologies related to image processing, the utilization of computer vision based on deep learning has increased. The algorithms for object recognition and detection based on deep learning required for image processing are diversified and advanced. Accordingly, problems that were difficult to solve with the existing methodology were solved more simply and easily by using deep learning. This paper introduces various deep learning-based object recognition and extraction algorithms used to detect and recognize various objects in an image and analyzes the technologies that attract attention.

• 정보관리학회지
• /
• 제35권2호
• /
• pp.63-88
• /
• 2018

본 연구는 LDA 토픽 모델과 딥 러닝을 적용한 단어 임베딩 기반의 Doc2Vec 기법을 활용하여 자질을 선정하고 자질집합의 크기와 종류 및 분류 알고리즘에 따른 분류 성능의 차이를 평가하였다. 또한 자질집합의 적절한 크기를 확인하고 문헌의 위치에 따라 종류를 다르게 구성하여 분류에 이용할 때 높은 성능을 나타내는 자질집합이 무엇인지 확인하였다. 마지막으로 딥 러닝을 활용한 실험에서는 학습 횟수와 문맥 추론 정보의 유무에 따른 분류 성능을 비교하였다. 실험문헌집단은 PMC에서 제공하는 생의학 학술문헌을 수집하고 질병 범주 체계에 따라 구분하여 Disease-35083을 구축하였다. 연구를 통하여 가장 높은 성능을 나타낸 자질집합의 종류와 크기를 확인하고 학습 시간에 효율성을 나타냄으로써 자질로의 확장 가능성을 가지는 자질집합을 제시하였다. 또한 딥 러닝과 기존 방법 간의 차이점을 비교하고 분류 환경에 따라 적합한 방법을 제안하였다.

• 스마트미디어저널
• /
• 제10권3호
• /
• 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.