• Journal of KIISE
• /
• v.45 no.2
• /
• pp.134-140
• /
• 2018

A chat system is a computer program that understands user's miscellaneous utterances and generates appropriate responses. Sometimes a chat system needs to answer users' simple information-seeking questions. However, previous generative chat systems do not consider how to embed knowledge entities (i.e., subjects and objects in triple knowledge), essential elements for question-answering. The previous chat models have a disadvantage that they generate same responses although knowledge entities in users' utterances are changed. To alleviate this problem, we propose a knowledge entity embedding method for improving question-answering accuracies of a generative chat system. The proposed method uses a Siamese recurrent neural network for embedding knowledge entities and their synonyms. For experiments, we implemented a sequence-to-sequence model in which subjects and predicates are encoded and objects are decoded. The proposed embedding method showed 12.48% higher accuracies than the conventional embedding method based on a convolutional neural network.

• Transactions on Semiconductor Engineering
• /
• v.1 no.1
• /
• pp.23-30
• /
• 2023

With the recent development of Generative AI technology by IT giants, the size of the transformer model is increasing exponentially over trillion won. In order to continuously enable these AI services, it is essential to reduce the weight of the model. In this paper, we find a hardware-friendly structured pruning pattern and propose a lightweight method of the transformer model. Since compression proceeds by utilizing the characteristics of the model algorithm, the size of the model can be reduced and performance can be maintained as much as possible. Experiments show that the structured pruning proposed when pruning GPT-2 and BERT language models shows almost similar performance to fine-grained pruning even in highly sparse regions. This approach reduces model parameters by 80% and allows hardware acceleration in structured form with 0.003% accuracy loss compared to fine-tuned pruning.

• KIPS Transactions on Software and Data Engineering
• /
• v.8 no.7
• /
• pp.311-316
• /
• 2019

Classification problem is to predict the class to which an input data belongs. One of the most popular methods to do this is training a machine learning algorithm using the given dataset. In this case, the dataset should have a well-balanced class distribution for the best performance. However, when the dataset has an imbalanced class distribution, its classification performance could be very poor. To overcome this problem, we propose an over-sampling scheme that balances the number of data by using Conditional Generative Adversarial Networks (CGAN). CGAN is a generative model developed from Generative Adversarial Networks (GAN), which can learn data characteristics and generate data that is similar to real data. Therefore, CGAN can generate data of a class which has a small number of data so that the problem induced by imbalanced class distribution can be mitigated, and classification performance can be improved. Experiments using actual collected data show that the over-sampling technique using CGAN is effective and that it is superior to existing over-sampling techniques.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.17 no.2
• /
• pp.291-298
• /
• 2022

A generative adversarial network (GAN) is a network in which two internal neural networks (generative network and discriminant network) learn while competing with each other. The generator creates an image close to reality, and the delimiter is programmed to better discriminate the image of the constructor. This technology is being used in various ways to create, transform, and restore the entire image X into another image Y. This paper describes a method that can be forged into another object naturally, after extracting only a partial image from the original image. First, a new image is created through the previously trained DCGAN model, after extracting only a partial image from the original image. The original image goes through a process of naturally combining with, after re-styling it to match the texture and size of the original image using the overall style transfer technique. Through this study, the user can naturally add/transform the desired object image to a specific part of the original image, so it can be used as another field of application for creating fake images.

• Design Convergence Study
• /
• v.15 no.6
• /
• pp.229-241
• /
• 2016

In generative design computer automatically and quickly generates many alternative design solutions, and the computer algorithms which perform the design tasks are important. The main purpose of this study is to propose a computer algorithm which generates three-dimensional shapes from a two-dimensional digital image such as a photograph or a painting. The base geometry of the final shape is a cylinder or sphere. A surface of the cylinder or sphere is deformed depending on the used image. The proposed algorithm was implemented as a computer program, and the program tested with several famous paintings. The algorithm and results presented in this study implicate the possibilities of the generative design which generates three-dimensional shapes from two-dimensional images. It is necessary to find and measure the values of the generated shape, and it will be a future research to find the relations of emotional and cognitive aspects between the input images and the generated shapes. Those studies are expected to expand the possibilities of generative design.

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.4
• /
• pp.31-38
• /
• 2024

This study proposes hyper-parameter settings for developing a generative AI-based learning support tool to facilitate programming education in online distance learning. We implemented an experimental tool that can set research hyper-parameters according to three different learning contexts, and evaluated the quality of responses from the generative AI using the tool. The experiment with the default hyper-parameter settings of the generative AI was used as the control group, and the experiment with the research hyper-parameters was used as the experimental group. The experiment results showed no significant difference between the two groups in the "Learning Support" context. However, in other two contexts ("Code Generation" and "Comment Generation"), it showed the average evaluation scores of the experimental group were found to be 11.6% points and 23% points higher than those of the control group respectively. Lastly, this study also observed that when the expected influence of response on learning motivation was presented in the 'system content', responses containing emotional support considering learning emotions were generated.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.8
• /
• pp.4255-4269
• /
• 2019

This paper proposes an image generation method using a Multi Discriminator Generative Adversarial Net (MDGAN) as a next generation 2D game sprite creation technique. The proposed GAN is an Autoencoder-based model that receives three areas of information-color, shape, and animation, and combines them into new images. This model consists of two encoders that extract color and shape from each image, and a decoder that takes all the values of each encoder and generates an animated image. We also suggest an image processing technique during the learning process to remove the noise of the generated images. The resulting images show that 2D sprites in games can be generated by independently learning the three image attributes of shape, color, and animation. The proposed system can increase the productivity of massive 2D image modification work during the game development process. The experimental results demonstrate that our MDGAN can be used for 2D image sprite generation and modification work with little manual cost.

• International journal of advanced smart convergence
• /
• v.12 no.2
• /
• pp.90-95
• /
• 2023

With the rapid development of deep learning and artificial intelligence, generative models have achieved remarkable success in the field of image generation. By combining the stable diffusion method with Web UI technology, a novel solution is provided for the application of AI painting generation. The application prospects of this technology are very broad and can be applied to multiple fields, such as digital art, concept design, game development, and more. Furthermore, the platform based on Web UI facilitates user operations, making the technology more easily applicable to practical scenarios. This paper introduces the basic principles of Stable Diffusion Web UI technology. This technique utilizes the stability of diffusion processes to improve the output quality of generative models. By gradually introducing noise during the generation process, the model can generate smoother and more coherent images. Additionally, the analysis of different model types and applications within Stable Diffusion Web UI provides creators with a more comprehensive understanding, offering valuable insights for fields such as artistic creation and design.

• Journal of the Korean Society for Industrial and Applied Mathematics
• /
• v.28 no.1
• /
• pp.1-21
• /
• 2024

This review comprehensively explores the evolution, theoretical underpinnings, variations, and applications of diffusion models. Originating as a generative framework, diffusion models have rapidly ascended to the forefront of machine learning research, owing to their exceptional capability, stability, and versatility. We dissect the core principles driving diffusion processes, elucidating their mathematical foundations and the mechanisms by which they iteratively refine noise into structured data. We highlight pivotal advancements and the integration of auxiliary techniques that have significantly enhanced their efficiency and stability. Variants such as bridges that broaden the applicability of diffusion models to wider domains are introduced. We put special emphasis on the ability of diffusion models as a crucial foundation model, with modalities ranging from image, 3D assets, and video. The role of diffusion models as a general foundation model leads to its versatility in many of the downstream tasks such as solving inverse problems and image editing. Through this review, we aim to provide a thorough and accessible compendium for both newcomers and seasoned researchers in the field.

• International Journal of Internet, Broadcasting and Communication
• /
• v.16 no.2
• /
• pp.81-86
• /
• 2024

Advancements in generative AI are reshaping graphic and 3D content design landscapes, where AI not only enriches graphic design but extends its reach to 3D content creation. Though 3D texture mapping through AI is advancing, AI-generated 3D modeling technology in this realm remains nascent. This paper presents AI 2D image-driven 3D modeling techniques, assessing their viability in 3D content design by scrutinizing various algorithms. Initially, four OBJ model-exporting AI algorithms are screened, and two are further evaluated. Results indicate that while AI-generated 3D models may not be directly usable, they effectively capture reference object structures, offering substantial time savings and enhanced design efficiency through manual refinements. This endeavor pioneers new avenues for 3D content creators, anticipating a dynamic fusion of AI and 3D design.