Digital Library[ Search Result ]
An Efficient User Interest Region Stitching Method using the RANSAC Algorithm
http://doi.org/10.5626/JOK.2019.46.10.1025
Recently, with the emergence of 5G technology capable of high-capacity wireless data transmission, technologies using 360-degree virtual reality images have attracted attention, and therefore, interest in image stitching is increasing. In this paper, we proposed a user"s interest region matching method using the RANSAC algorithm. The proposed method uses the RANSAC algorithm and assigns a high weight to the region of interest selected by the user to perform image matching. It can be selectively performed in regions requiring natural stitching. The corresponding points included in the region of interest are set to have high weight and are necessarily included in the sample selection of the RANSAC algorithm. Additionally, the degree of matching to a specific region can be adjusted depending on whether several feature points are included. The method of interest region matching consists of setting the region of interest, increasing the weights of the corresponding points in the region of interest, creating the model using the RANSAC algorithm, and setting the inliers and outliers of the corresponding points using the model. The results of this study confirmed that users can get results approximating reality by performing matching based on points corresponding to selected regions of interest.
PARPA: A Parallel Framework Simultaneously Using Heterogeneous Architecture for High Performance Computing
Hyojae Cho, Taehyun Han, Hyeonmyeong Lee, Heeseung Jo
http://doi.org/10.5626/JOK.2019.46.9.876
With the substantial performance improvements achieved in GPU, they have come to be commonly used not only in computer graphics but also in high performance computing. Simply using a CPU and a GPU concurrently is not difficult. However, distributing works and adjusting the computing ratio among these heterogeneous processors are challenging issues. We propose a novel framework in this paper, named PARPA, which automatically distributes and processes tasks to a CPU and a GPU. PARPA can maximize computation performance by using a CPU and a GPU simultaneously. The load balancing between them can be performed dynamically based on their usage and features. The evaluation results indicate that PARPA shows 3.48 times better performance.
Effective Generative Chatbot Model Trainable with a Small Dialogue Corpus
Jintae Kim, Hyeon-gu Lee, Harksoo Kim
http://doi.org/10.5626/JOK.2019.46.3.246
Contrary to popular retrieval-based chatbot models, generative chatbot models do not depend on predefined responses, but rather generate new responses based on well-trained neural networks. However, they require a large number of training corpus in the form of query-response pairs. If the training corpus are insufficient, they make grammatical errors emanating from out-of-vocabulary or sparse data problems, mostly in longer sentences. To overcome this challenge, we proposed a chatbot model based on sequence-to-sequence neural network using a mixture of words and syllables as encoding-decoding units. Moreover, we proposed a two-step training procedure involving pre-training using a large non-dialogue corpus and retraining using a smaller dialogue corpus. In the experiment involving small dialogue corpus (47,089 query-response pairs for training and 3,000 query-response pairs for evaluation), the proposed encoding-decoding units resulted to a reduction in out-of-vocabulary problem while the two-step training method led to improved performance measures like BLEU and ROUGE.
Korean Machine Reading Comprehension using Reinforcement Learning and Dual Co-Attention Mechanism
http://doi.org/10.5626/JOK.2018.45.9.932
Machine Reading Comprehension is a question-answering model for the purposes of understanding a given document and then finding the correct answer within the document. Previous studies on the Machine Reading Comprehension model have been based on end-to-end neural network models with various attention mechanisms. However, in the previous models, difficulties arose when attempting to find answers with long dependencies between lexical clues because these models did not use grammatical and syntactic information. To resolve this problem, we propose a Machine Reading Comprehension model with a dual co-attention mechanism reflecting part-of-speech information and shortest dependency path information. In addition, to increase the performances, we propose a reinforce learning method using F1-scores of answer extraction as rewards. In the experiments with 18,863 question-answering pairs, the proposed model showed higher performances (exact match: 0.4566, F1-score: 0.7290) than the representative previous model.
Search
Journal of KIISE
- ISSN : 2383-630X(Print)
- ISSN : 2383-6296(Electronic)
- KCI Accredited Journal
Editorial Office
- Tel. +82-2-588-9240
- Fax. +82-2-521-1352
- E-mail. chwoo@kiise.or.kr