Earticle

Recently, research on pedestrian action recognition from the vehicle’s viewpoint is being studied in many ways. The information about pedestrian action classification is very important for autonomous driving to determine safe path planning and avoid accidents. To provide a computationally efficient solution to pedestrian action recognition, this paper proposes a multi-head CNN model to currently extract multi-actions of pedestrians from the unified model. This model consists of one pre-trained backbone network and two head networks. One head network classifies Gait (walking/standing) and the second classifies Attention (looking/non-looking) of pedestrians. The proposed model offers a lighter model with smaller memory, faster processing speed, and alleviates data imbalance problem – a common problem found in most of dataset – leading to improved accuracy.

Federated learning provides an efficient integrated model for distributed data, allowing the local training of different data. Meanwhile, the goal of multi-task learning is to simultaneously establish models for multiple related tasks, and to obtain the underlying main structure. However, traditional federated multi-task learning models not only have strict requirements for the data distribution, but also demand large amounts of calculation and have slow convergence, which hindered their promotion in many fields. In our work, we apply the rank constraint on weight vectors of the multi-task learning model to adaptively adjust the task's similarity learning, according to the distribution of federal node data. The proposed model has a general framework for solving optimal solutions, which can be used to deal with various data types. Experiments show that our model has achieved the best results in different dataset. Notably, our model can still obtain stable results in datasets with large distribution differences. In addition, compared with traditional federated multi-task learning models, our algorithm is able to converge on a local optimal solution within limited training iterations.

본 논문에서는 UD Korean Kaist v2.3 코퍼스를 이용하여 범용 품사 태깅, 표제어추출 그리고 의존 구문분석을 동시에 예측할 수 있는 보편적 다중 작업 모델을 제안하였다. 제안 모델은 사전학습 언어모델인 다국어 BERT (Multilingual BERT)와 한국어 BERT (KR-BERT와 KoBERT)을 대상으로 추가학습 (fine-tuning)을 수행하여 BERT 모델의 자가-집중 (self-attention) 기법과 그래프 기반 Biaffine attention 기법을 적용하여 제안 모델의 성능을 비교 분석하였다.

In this paper, we propose a multi-task model that can simultaneously predict general-purpose tasks such as part-of-speech tagging, lemmatization, and dependency parsing using the UD Korean Kaist v2.3 corpus. The proposed model thus applies the self-attention technique of the BERT model and the graph-based Biaffine attention technique by fine-tuning the multilingual BERT and the two Korean-specific BERTs such as KR-BERT and KoBERT. The performances of the proposed model are compared and analyzed using the multilingual version of BERT and the two Korean-specific BERT language models.

This study investigates how model text feedback affects Korean EFL learners' subsequent writing and their uptake of targeted grammatical structures, extending Kim and Park's (2025) research on noticing patterns in second language (L2) writing. To examine both general learning transfer and specific grammatical acquisition, this study focused on two complex conditional structures that pose persistent challenges for Korean EFL learners. Forty-four intermediate-level nursing students completed a three-stage writing task targeting these previously taught structures: second conditional and should have/would have constructions. While the treatment group (n=22) received a native-speaker model text during revision, the control group (n=21) revised independently. The results reveal substantial learning effects. Problematic features noticed (PFNs) declined 61% in the treatment group (220 to 86 instances) versus 52% in the control group (210 to 100), though fundamental noticing patterns remained stable. The treatment group retained more features from revision (36 features versus 11), and their target structure accuracy improved dramatically: second conditional rose from 40.91% to 78.79% (p=.002) and should have/would have from 27.27% to 62.50% (p=.003), accompanied by increased attempt frequency. Control group improvements were modest and statistically non-significant. Qualitative analysis uncovered an intriguing tension: some participants felt creatively constrained despite objective improvements. The persistence of similar noticing patterns (lexical 36%, grammatical 27%, ideas/expression 27%) across stages, despite model exposure, suggests that while performance improves, fundamental cognitive processes during composition remain stable. These findings demonstrate that model texts powerfully enhance target structure acquisition and feature retention, functioning as effective tools for grammatical development without disrupting learners' autonomous problem-identification. The study illuminates how model texts facilitate learning transfer from revision to independent writing, confirming their value extends beyond immediate revision to sustainable development.

딥러닝 기반 자율 주행 기술이 발전함에 따라 다양한 목적의 인공지능 모델이 연구되었다. 연구된 여러 모델들을 동시에 구동하여 자율주행 시스템을 개발한다. 그러나 동시에 인공지능 모델을 사용하면서 많은 하드웨어 자원 소비가 증가한다. 이를 해결하기 위해 본 논문은 백본 모델을 공유하며 다중 태스크를 고속으로 수행할 수 있는 Multi-Task CNN 모델을 제안한다. 이를 통해 AI모델을 사용하기 위한 백본 수의 증가를 해결할 수 있었습니다. 제안하는 CNN 모델은 기존 모델 대비 50% 이상 웨이트 파라미터 수를 감소시키며, 3배 이상의 FPS 속도를 향상시켰다. 또한, 차선인식은 Instance segmentation 기반으로 차선검출 및 차선별 Labeling을 모두 출력한다. 그러나 기존 모델에 비해 정확도가 감소하는 부분에 대해서는 추가적인 연구가 필요하다.

As deep learning-based autonomous driving technology develops, artificial intelligence models for various purposes have been studied. Based on these studies, several models were used simultaneously to develop autonomous driving systems. It can occur by increasing hardware resource consumption. We propose a multi-tasks model using a shared backbone to solve this problem. This can solve the increase in the number of backbones for using AI models. As a result, in the proposed lightweight model, the model parameters could be reduced by more than 50% compared to the existing model, and the speed could be improved. In addition, each lane can be classified through lane detection using the instance segmentation method. However, further research is needed on the decrease in accuracy compared to the existing model.

RBAC은 역할 계층구조에서 권한의 계승과 의무분리와 같은 제약조건을 다룸으로써 접근권한의 관리를 수월하게 하고 기업환경을 잘 반영말 수 있는 장점이 있다. 하지만 RBAC은 현실세계의 기업환경에서 빈번히 이루어지는 권한의 위임을 제대로 구현하지 못한다는 문제점을 가지고 있다. 본 논문에서는 자신의 고유역할 뿐만 아니라 상위 역할로부터 위임받은 새로운 위임 역할을 생성함으로써 역할계층 구조상의 다른 역할의 사용자에게 다른 과업을 할당해 줄 수 있도록 하여, 최소 권한의 원칙을 만족하는 다단계 위임을 구현하였다. 위임 시에 생길 수 있는 보안 문제를 해결하기 위해서 역할단위가 아닌 과업단위의 위임으로 제안하고, 과업단위의 의무분리를 적용하였으며, 위임 할 수 있는 과업을 규정하고 최하위 역할을 지정하였다. 기존의 다단계 모델에서 제안된 기법과의 비교를 통해서 본 논문에서 제안된 기법이 실제 기업에서 이루어지는 다단계 위임을 타당하게 구현할 수 있음을 보인다. 또한 T-RBCA을 기반으로 ARBAC97을 적용해서 제안된 기법을 모델링하고 Prototype을 구현하였다.