전체 글83 삼성전자 DS 부문 박사대여장학생 합격 2025년도 삼성전자 DS 부문 박사대여장학생 합격 후기를 이제서야 올려본다.작년에는 너무 정신이 없었다 :( 24년도에도 관심이 있어서 교내에서 열린 박사대여장학생 설명회를 들으러 갔었다.당시 박사대여장학생은 박사 과정은 입학한 당해(1년차)부터, 통합 과정은 3년차부터 지원이 가능하다고 해서 지원하지 못했다. ㅠ 그래서 3년차가 된 2025년 4월에 바로 지원했다. 작년에는 박사대여장학생 TO가 적어서 서류 탈락도 많을 거고, 회사에서는 최대한 빨리 활용 가능한 인력을 선호하기에 저년차는 상대적으로 더 안 뽑으실거라는 소문을 들어서 최선을 다하더라도 떨어질 수 있다고 생각했고 큰 기대는 안했다.+ 떨어지더라도 저년차라 그 다음 기회를 노려도 된다고 생각했다. 그런데 서류가 붙어서 운좋게 면접을 들어.. 2026. 6. 20. [논문 리뷰] TIMER-XL: LONG-CONTEXT TRANSFORMERS FORUNIFIED TIME SERIES FORECASTING (ICLR, 2025) https://arxiv.org/abs/2410.04803 1. Introduction이전에 리뷰한 MOIRAI처럼 Time series forecasting을 위한 Foundation Model을 연구한 논문이다.이번에는 Decoder-only transformer 구조다. ( MOIRAI는 Encoder-only) 저자들은 Transformers for language and vision은 훨씬 긴 Token context를 활용하는 반면, Time series transformers는 수백 개 이하의 patch token만 보는 짧은 context만을 다룬다는 문제를 제기한다. 이 연구에서 다루고자 하는 핵심 research gap은 세 가지다.첫째, Univariate에서는 짧은 context는 .. 2026. 6. 10. [논문 리뷰] Unified Training of Universal Time Series Forecasting Transformers (ICML, 2024) https://arxiv.org/abs/2402.02592 Unified Training of Universal Time Series Forecasting TransformersDeep learning for time series forecasting has traditionally operated within a one-model-per-dataset framework, limiting its potential to leverage the game-changing impact of large pre-trained models. The concept of universal forecasting, emerging from pre-arxiv.org 1. IntroductionTime series forec.. 2026. 6. 10. WSL2 경로 Windows 탐색기로 여는 방법 우선 wsl을 실행시켜준다. cd {원하는 경로}를 통해 원하는 경로로 이동해준다. 이후, 아래 명령어를 입력하면 된다. (GPT 및 다른 블로그를 보면 이 방법을 가장 많이 사용)explorer.exe . 위 방법이 안되는 경우내 경우에서는 계속 기본 '문서' 페이지만 열리고, 해당 경로가 열리지 않았다. 다양한 이유가 있다고 한다.. 시도해본 다른 방법은 두가지다. 1. wslpath -w "$PWD" 를 입력하면 현재 위치를 Windows 기준 경로로 변환해준다.이를 활용해서 아래 명령어를 입력하면 Windows 탐색기에서 해당 경로를 열 수 있다.explorer.exe "$(wslpath -w "$PWD")" 2. 난 위 방법 모두 실패했고 이 방법으로 열 수 있었다.마찬가지로 원하는 경로로 이.. 2026. 5. 4. [논문 리뷰] Advances in Multimodal Adaptation and Generalization: From Traditional Approaches to Foundation Models-2 (IEEE TPAMI, 2026) Paper: https://ieeexplore.ieee.org/document/11342305 Advances in Multimodal Adaptation and Generalization: From Traditional Approaches to Foundation ModelsDomain adaptation and generalization are crucial for real-world applications, such as autonomous driving and medical imaging where the model must operate reliably across environments with distinct data distributions. However, these tasks are c.. 2026. 4. 22. [논문 리뷰] Advances in Multimodal Adaptation and Generalization: From Traditional Approaches to Foundation Models-1 (IEEE TPAMI, 2026) Paper: https://ieeexplore.ieee.org/document/11342305 Advances in Multimodal Adaptation and Generalization: From Traditional Approaches to Foundation ModelsDomain adaptation and generalization are crucial for real-world applications, such as autonomous driving and medical imaging where the model must operate reliably across environments with distinct data distributions. However, these tasks are c.. 2026. 4. 22. [논문 리뷰] Extremely simple multimodal outlier synthesis for out-of-distribution detection and segmentation (NeurIPS, 2025) Paper: https://arxiv.org/abs/2505.16985 Extremely Simple Multimodal Outlier Synthesis for Out-of-Distribution Detection and SegmentationOut-of-distribution (OOD) detection and segmentation are crucial for deploying machine learning models in safety-critical applications such as autonomous driving and robot-assisted surgery. While prior research has primarily focused on unimodal image data,arxiv... 2026. 4. 18. [논문 리뷰] Few-Shot Anomaly Detection via Category-Agnostic Registration Learning (IEEE TNNLS, 2024) Paper: https://arxiv.org/abs/2406.08810 Few-Shot Anomaly Detection via Category-Agnostic Registration LearningMost existing anomaly detection (AD) methods require a dedicated model for each category. Such a paradigm, despite its promising results, is computationally expensive and inefficient, thereby failing to meet the requirements for realworld applications. Insarxiv.org Background일반적인 Unsuper.. 2026. 4. 15. [논문 리뷰] Source-Free Domain Adaptation with Frozen Multimodal Foundation Model (CVPR, 2024) Paper: https://arxiv.org/abs/2311.16510 Source-Free Domain Adaptation with Frozen Multimodal Foundation ModelSource-Free Domain Adaptation (SFDA) aims to adapt a source model for a target domain, with only access to unlabeled target training data and the source model pre-trained on a supervised source domain. Relying on pseudo labeling and/or auxiliary supervisioarxiv.org Introduction기존 Unsuperv.. 2026. 2. 21. [논문 리뷰] NVIDIA C-RADIOv4 (Tech Report, 2026) Report: https://arxiv.org/abs/2601.17237 C-RADIOv4 (Tech Report)By leveraging multi-teacher distillation, agglomerative vision backbones provide a unified student model that retains and improves the distinct capabilities of multiple teachers. In this tech report, we describe the most recent release of the C-RADIO familarxiv.org 엔비디아가 강력한 성능과 효율성을 겸비한 새로운 Vision Foundation Model을 공개했다.다른 빅테크 기업이 .. 2026. 1. 28. [논문 리뷰] BLIP: Bootstrapping Language-Image Pre-training forUnified Vision-Language Understanding and Generation (ICML, 2022) Paper: https://arxiv.org/abs/2201.12086 BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and GenerationVision-Language Pre-training (VLP) has advanced the performance for many vision-language tasks. However, most existing pre-trained models only excel in either understanding-based tasks or generation-based tasks. Furthermore, performance improvement has b.. 2025. 9. 19. [논문 리뷰] SigLIP 2: Multilingual Vision-Language Encoderswith Improved Semantic Understanding,Localization, and Dense Features (2025) Paper: https://arxiv.org/abs/2502.14786 SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense FeaturesWe introduce SigLIP 2, a family of new multilingual vision-language encoders that build on the success of the original SigLIP. In this second iteration, we extend the original image-text training objective with several prior, independently.. 2025. 9. 19. 이전 1 2 3 4 ··· 7 다음
