λ°μ΄ν° μ€μ¬ AIλ λ°μ΄ν° νμ§ ν₯μκ³Ό ν΅μ¬ λ°μ΄ν° μ μ μ ν΅ν΄ νμ΅ μκ³ λ¦¬μ¦ μ±λ₯μ μ΅μ ννλ μ°κ΅¬ λΆμΌμ λλ€. μ΄λ₯Ό ν΅ν΄ ν¨μ¨μ μ΄κ³ κ°κ±΄ν λͺ¨λΈ κ°λ°μ λͺ©νλ‘ ν©λλ€.Β
Data-centric AI is a research field focused on optimizing the performance of learning algorithms by improving data quality and selecting key data. The goal is to develop efficient and robust models.Β
Keywords: Active Learning, Noise Label Learning, Continual Learning, Semi-supervised Learning, Coreset Selection
μλ£ μμ λΆμμ μ»΄ν¨ν° λΉμ κ³Ό μμ± λͺ¨λΈμ λΉλ‘―ν λ€μν λ¨Έμ λ¬λ κΈ°μ μ νμ©νμ¬ λ°©λν μλ£ μμ λ°μ΄ν°λ₯Ό μ¬μΈ΅ λΆμν©λλ€. μλ£ μμ λΆμμ ν΅ν΄ μ§λ³ μ§λ¨μ μ νμ±μ λμ΄λ©° μΉλ£ κ³νμ μ΅μ ννλ κ²μ λͺ©νλ‘ ν©λλ€.
Medical image analysis is a research field that aims to deeply analyze vast amounts of medical imaging data using various machine learning techniques, including computer vision and generative models. Through this, it seeks to enhance the accuracy of disease diagnosis and optimize treatment planning.Β
Keywords: Multiple Instance Learning, Image Segmentation, Object Detection, Generative Model, Image Reconstruction, LLM
λ°μ΄μ€ AI μ°κ΅¬λ μ μ½ ν보 λ¬Όμ§ λ°κ΅΄κ³Ό λ¨λ°±μ§-λΆμ μνΈμμ© λΆμμ κ³ λννκ³ , μμ± λͺ¨λΈμ νμ©νμ¬ ν¨λ₯μ΄ λ°μ΄λκ³ ν©μ± κ°λ₯μ±μ΄ λμ λΆμ μ€κ³λ₯Ό μλνν¨μΌλ‘μ¨, ν¨μ¨μ μΈ μ μ½ κ°λ° νλ‘μΈμ€λ₯Ό ꡬννκ³ μλ‘μ΄ μΉλ£μ λ°κ΅΄μ κΈ°μ¬ν©λλ€.Β
AI Bio research enhances drug candidate discovery and protein-molecule interaction analysis, automating molecular design using generative models to create compounds with high efficacy and synthetic feasibility. This process enables a more efficient drug development workflow and contributes to the discovery of new therapeutic agents.Β
Keywords: Drug Discovery, Generative Model,Β LLM, RAG, PEFT, Drug Target Interaction, Reinforcement Learning
μ°ν©νμ΅μ λ°μ΄ν°λ₯Ό μ€μ μλ²λ‘ μ μ‘νμ§ μκ³ κ° λλ°μ΄μ€μμ λΆμ° λ°©μμΌλ‘ λͺ¨λΈμ νμ΅νμ¬ κ°μΈμ 보 보νΈμ λ°μ΄ν° 보μμ κ°ννλ νλ μμν¬μ λλ€. μ°λ¦¬ μ°κ΅¬λ μ°ν©νμ΅ κ³Όμ μμ λ°μνλ λ€μν λ¬Έμ λ₯Ό ν΄κ²°νκ³ , μ°Έμ¬μλ€μ΄ μ§μμ μΌλ‘ νμ΅μ μ°Έμ¬νλλ‘ μ λνλ μκ³ λ¦¬μ¦μ κ°λ°ν©λλ€.Β
Federated learning is a framework that enhances privacy and data security by training models in a decentralized manner on individual devices, without sending data to a central server. Our research focuses on addressing various challenges that arise during the federated learning process and developing algorithms that encourage participants to continue engaging in the learning process.Β
Keywords: Federated Learning, Personalization, Incentive Mechanism. Backdoor Attack, Multi-agent Learning