• Annual Conference of KIPS
• /
• 2024.05a
• /
• pp.580-581
• /
• 2024

고령화 사회에 접어들면서 황반 변성과 당뇨 망막 병증 등 시야결손을 동반하는 안구 질환의 발병률은 증가하지만 이러한 질환의 조기 발견에 인공지능을 접목시킨 연구는 부족한 실정이다. 본 논문은 안구 질환 자가 검사용 인공 신경망을 학습시키기 위한 데이터 베이스 구축 방법을 제안한다. MNIST와 CIFAR-10을 합성하여 중첩 이미지 데이터셋인 G-Dataset을 생성하였고, 7개의 인공신경망에 학습시켜 최종적으로 90% 이상의 정확도를 얻음으로 그 유효성을 입증하였다. G-Dataset을 안구 질환 자가 검사용 딥러닝 모델에 학습시켜 모바일 어플에 적용하면 사용자가 주기적인 검사를 통해 안구 질환을 조기에 진단하고 치료할 수 있을 것으로 기대된다.

• Annual Conference of KIPS
• /
• 2017.11a
• /
• pp.807-809
• /
• 2017

Generative adversarial networks (GAN)에 대한 간략하게 설명하고, MNIST (숫자 손 글씨 데이터 셋)를 이용한 간단한 실험을 통해 GAN 구조 구조의 이해를 돕는다. 그리고 GAN이 어떻게 응용이 되고있는지 다양한 논문들을 통해 살펴본다. 본 고에서는 GAN 논문들을 크게 이미지 스타일 변경, 3D 오브젝트 추정, 손상된 이미지 복원, 언어의 시각화, 기타 등으로 분류하였다.

• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.9
• /
• pp.375-384
• /
• 2021

Deep learning techniques have been proven to have high performance in image processing and are applied in various fields. The most widely used methods for validating a deep learning model include a holdout verification method, a k-fold cross verification method, and a bootstrap method. These legacy methods consider the balance of the ratio between classes in the process of dividing the data set, but do not consider the ratio of various features that exist within the same class. If these features are not considered, verification results may be biased toward some features. Therefore, we propose a deep learning model validation method based on data feature coverage for image classification by improving the legacy methods. The proposed technique proposes a data feature coverage that can be measured numerically how much the training data set for training and validation of the deep learning model and the evaluation data set reflects the features of the entire data set. In this method, the data set can be divided by ensuring coverage to include all features of the entire data set, and the evaluation result of the model can be analyzed in units of feature clusters. As a result, by providing feature cluster information for the evaluation result of the trained model, feature information of data that affects the trained model can be provided.

• Annual Conference of KIPS
• /
• 2024.05a
• /
• pp.207-209
• /
• 2024

딥러닝은 선형 연산과 비선형 연산을 조합하여 목표로 하는 시스템을 잘 표현할 수 있는 함수를 찾기 위해 사용하며, 이미지 분류 및 생성, 거대 언어 모델 및 객체 인식의 영역에서 활발하게 사용되고 있다. 그러나 딥러닝 연산을 위해서는 모델과, 연산을 수행하고자 하는 데이터가 하나의 공간에 저장되어야 한다. 모델과 데이터를 데이터 소유자가 관리할 경우, 데이터 소유자가 모델 데이터의 프라이버시를 침해할 수 있으며, 이는 모델을 적대적 예제 생성 공격에 취약하도록 만드는 원인이 된다. 한편 모델과 데이터를 모델 소유자가 관리할 경우, 모델 소유자는 데이터의 프라이버시를 침해하여 데이터 소유자의 정보를 악의적으로 이용할 수 있다. 본 논문에서는 딥러닝 모델과 데이터의 프라이버시를 모두 보호하기 위해 주어진 딥러닝 모델의 암호화와 복호화를 수행하는 EncNet 을 구현하였으며, MNIST 와 Cifat-10 데이터셋에 대하여 실효성을 테스트하였다.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.24 no.1
• /
• pp.181-187
• /
• 2024

There are two kinds of well-known CNN methods, the group equivariant CNN and the CNN using steerable filters, which have excellent classification performances for randomly rotated objects in image space. This paper describes their mathematical structures and introduces implementation methods. We implement them, including the existing CNN, which have the same number of filters, then compare and analyze their performances by simulating them with the randomly rotated MNIST. According to the experimental results, the steerable CNN, which shows a classification improvement over the others, has a relatively small number of parameters to learn, so performance degradation is relatively small even when the size of the training dataset is reduced.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.21 no.2
• /
• pp.189-194
• /
• 2020

This paper analyzes the gradient descent method, which is the one most used for learning neural networks. Learning means updating a parameter so the loss function is at its minimum. The loss function quantifies the difference between actual and predicted values. The gradient descent method uses the slope of the loss function to update the parameter to minimize error, and is currently used in libraries that provide the best deep learning algorithms. However, these algorithms are provided in the form of a black box, making it difficult to identify the advantages and disadvantages of various gradient descent methods. This paper analyzes the characteristics of the stochastic gradient descent method, the momentum method, the AdaGrad method, and the Adadelta method, which are currently used gradient descent methods. The experimental data used a modified National Institute of Standards and Technology (MNIST) data set that is widely used to verify neural networks. The hidden layer consists of two layers: the first with 500 neurons, and the second with 300. The activation function of the output layer is the softmax function, and the rectified linear unit function is used for the remaining input and hidden layers. The loss function uses cross-entropy error.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.17 no.4
• /
• pp.229-235
• /
• 2024

Recent years have seen active research on methods for efficiently processing and interpreting large volumes of data in the fields of artificial intelligence and machine learning. One of these data processing technologies, Vector Symbolic Architecture (VSA), offers an innovative approach to representing complex symbols and data using high-dimensional vectors. VSA has garnered particular attention in various applications such as natural language processing, image recognition, and robotics. This study quantitatively evaluates the characteristics and performance of VSA methodologies by applying five VSA methodologies to the MNIST dataset and measuring key performance indicators such as encoding speed, decoding speed, memory usage, and recovery accuracy across different vector lengths. BSC and VT demonstrated relatively fast performance in encoding and decoding speeds, while MAP and HRR were relatively slow. In terms of memory usage, BSC was the most efficient, whereas MAP used the most memory. The recovery accuracy was highest for MAP and lowest for BSC. The results of this study provide a basis for selecting appropriate VSA methodologies depending on the application area.

• Journal of Platform Technology
• /
• v.6 no.4
• /
• pp.34-40
• /
• 2018

Currently, artificial neural networks perform well for a single task, but NN have the problem of forgetting previous learning by learning other kinds of tasks. This is called catastrophic forgetting. To use of artificial neural networks in general purpose this should be solved. There are many efforts to overcome catastrophic forgetting. However, even though there was a lot of effort, it did not completely overcome the catastrophic forgetting. In this paper, we propose sequential iterative learning using core concepts used in elastic weight consolidation (EWC). The experiment was performed to reproduce catastrophic forgetting phenomenon using EMNIST data set which extended MNIST, which is widely used for artificial neural network learning, and overcome it through sequential iterative learning.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.32 no.5
• /
• pp.817-826
• /
• 2022

AI technology is being successfully introduced in many fields, and models deployed as a service are deployed with black box environment that does not expose the model's information to protect intellectual property rights and data. In a black box environment, attackers try to steal data or parameters used during training by using model output. This paper proposes a method of inferring the type of model to directly find out the composition of layer of the target model, based on the fact that there is no attack to infer the information about the type of model from the deep learning model. With ResNet, VGGNet, AlexNet, and simple convolutional neural network models trained with MNIST datasets, we show that the types of models can be inferred using the output values in the gray box and black box environments of the each model. In addition, we inferred the type of model with approximately 83% accuracy in the black box environment if we train the big and small relationship feature that proposed in this paper together, the results show that the model type can be infrerred even in situations where only partial information is given to attackers, not raw probability vectors.

• The Journal of the Convergence on Culture Technology
• /
• v.3 no.4
• /
• pp.165-169
• /
• 2017

CNN(Convolutional Nerual Network) is one of the algorithms that show superior performance in image recognition and classification among machine learning algorithms. CNN is simple, but it has a large amount of computation and it takes a lot of time. Consequently, in this paper we performed an parallel processing unit for the convolution layer, pooling layer and the fully connected layer, which consumes a lot of handling time in the process of CNN, through the SIMT(Single Instruction Multiple Thread)'s structure of GPGPU(General-Purpose computing on Graphics Processing Units).And we also expect to improve performance by reducing the number of memory accesses and directly using the output of convolution layer not storing it in pooling layer. In this paper, we use MNIST dataset to verify this experiment and confirm that the proposed CNN structure is 12.38% better than existing structure.