CIDA LabUniversity of Seoul

About

CIDA Lab group photo at KCC 2026

Our research centers on neuro-symbolic algorithms, formal verification of neural networks, next-generation neural architectures & learning algorithms, and multi-agent reinforcement learning — which we apply to real-world domains such as sports data analytics and AI-driven program understanding & generation. Meet our team members, browse our publications, or visit Prof. Sang-Ki Ko's personal website for more details.

모집 중 / Recruiting 대학원생 · 학부 연구생

CIDA Lab에서 함께 연구할 학생을 모집합니다

저희 연구실은 뉴로-심볼릭 알고리즘, 신경망 정형 검증, 차세대 신경망 구조·학습 알고리즘, 다중 에이전트 강화학습을 핵심 주제로 연구하며, 이를 스포츠 데이터 분석과 AI 기반 프로그램 이해·생성 같은 실제 문제에 응용하고 있습니다. AAAI, KDD, IJCAI, CIKM, ACL, MIT Sloan Sports Analytics Conference 등 세계 최고 수준의 학회에서 꾸준히 성과를 내고 있습니다. 이러한 주제에 관심이 있고 대학원 진학을 고려 중인 학생은 아래 메일로 연락 주시기 바랍니다.

  • 학부 연구생으로 지원하고자 하는 경우, 메일에 성적표를 첨부해 주세요.
  • 학부 연구생은 최소 1년 이상 연구실에서 활동할 의지가 있는 학생만 지원해 주시기 바랍니다.
sangkiko@uos.ac.kr 로 문의하기

Research Areas

Symbolic × Neural

Neuro-Symbolic Algorithms

Combining symbolic, automata-theoretic methods with neural learning: regular expression synthesis from examples, regular language inference, descriptional complexity, and Simon's congruence.

 Safety & Correctness

Formal Verification of Neural Networks

Safety and correctness verification of DNNs and spiking neural networks (SNNs) using automata-theoretic and model-checking techniques.

 Beyond Deep Networks

Next-Generation Neural Architectures & Learning

Spiking neural networks and biologically inspired learning algorithms such as STDP, with a focus on training efficiency and adversarial robustness.

 Multi-Agent Systems

Multi-Agent Reinforcement Learning

Reinforcement learning for multiple interacting agents, game-playing agents, and strategy optimization in dynamic, uncertain environments.
Real-World Impact

ML/AI Applications

Applying cutting-edge machine learning and AI to real-world domains, built on the methods above. Two flagship areas:

Sports Data Analytics Player performance evaluation, multi-agent trajectory inference, formation analysis, strategy optimization Program Understanding & Generation Program repair, time complexity prediction, grammar-based test generation, LLM-based code analysis

Read more about our research

Recent Highlights

  • ECML PKDD 2026 KIISE CS Top Conf · ADS Track

    ScoutGPT: a generative Transformer that models matches as language for counterfactual player valuation in football — led by Miru Hong.

  • IJCAI 2026 BK21 Top Conf · IF 4

    ReSyn: a generalized recursive regular expression synthesis framework — led by Seongmin Kim.

  • MIT Sloan SAC 2026 Finalist · Top 7 / 200+

    "Valuing La Pausa" — selected as a finalist at the MIT Sloan Sports Analytics Conference.

  • IJCAI 2025 BK21 Top Conf · IF 4

    LogiCase: effective test case generation from logical descriptions.

  • EMNLP 2025 BK21 Top Conf · IF 3

    CodeComplex: benchmark dataset for worst-case time complexity prediction.

  • CIKM 2025 BK21 Top Conf · IF 3

    Multi-agent trajectory imputation in soccer from event and snapshot data — led by Geonhee Jo & Miru Hong.

  • ECML PKDD 2025 KIISE CS Top Conf

    Trajectory imputation with derivative-accumulating self-ensemble — led by Han-Jun Choi.

  • MIT Sloan SAC 2025

    exPress: contextual player valuation in pressing situations.

View all publications

Latest News

View all news

Search