The Laboratory of Big Data Analytics and Learning (BigDyL pronounced as Big Deal) in Web, Industrial, and Scientific Environments (WISE), was established in March 2020 and is now studying the following problems:
Time Series/Spatiotemporal Synthesis, Forecasting, and Anomaly Detection,
Recommendation (Collaborative Filtering and Sequential Recommendation),
Tabular Data Synthesis and Representation Learning,
Differential Equation-based Deep Learning and Physics-informed Neural Networks,
Graph Convolutional Networks,
Deep Generative Models.
We have one open position for PhD, starting Spring 2025, for the following topics: Physics-informed Neural Networks, Mesh Graph Networks, Neural Operators. We strongly recommend that you do an internship for this position if interested.Β
We tentatively do not have other open positions.
π° News
(May 7, 2024) Prof. Noseong Park will chair the tutorial session of the 18th ACM International Web Search and Data Mining Conference (WSDM).
(May 2, 2024) Four papers were accepted in International Conference on Machine Learning (ICML) 2024.
(April 17, 2024) One paper, titled "SVD-AE: Simple Autoencoders for Collaborative Filtering," was accepted in International Joint Conference on Artificial Intelligence (IJCAI) 2024.
(March 4, 2024) Two papers, titled "Extension of Physics-informed Neural Networks to Solving Parameterized PDEs," and "SPI-GAN: Denoising Diffusion GANs with Straight-Path Interpolations," were accepted in ICLR Workshops 2024 .
(February 1, 2024) BigDyL in WISE moved to the School of Computing in KAIST, Daejeon. Prof. Noseong Park was appointed as a tenured associate professor at KAIST. [νκ³Ό 곡μ§]
(January 16, 2024) Two papers, titled "PAC-FNO: Parallel-Structured All-Component Fourier Neural Operators for Recognizing Low-Quality Images," and "Learning Flexible Body Collision Dynamics with Hierarchical Contact Mesh Transformer," were accepted in International Conference on Learning Representations (ICLR) 2024.
(January 11, 2024) κΈμ΅μμν μ£Όκ΄ D-ν μ€νΈλ² λ μ¬μ μμ μμ κ· νμμ΄ λ₯λ¬λγλ¨Έμ λ¬λ νμ© κΈμ΅μ¬κΈ°νμ§ λͺ¨λΈμ μ΅μ ν μ°κ΅¬λ‘ μ΅μ°μμ(νν ν¬μ§μμΌν° μ΄μ¬μ₯μ)μ μμνμμ΅λλ€. [보λμλ£]
(December 9, 2023) Two papers, titled "Operator-learning-inspired Modeling of Neural Ordinary Differential Equations," and "An Attentive Inductive Bias for Sequential Recommendation beyond the Self-Attention," were accepted in AAAI Conference on Artificial Intelligence (AAAI) 2024.
(October 29, 2023) One paper, titled "Long-term Time Series Forecasting based on Decomposition and Neural Ordinary Differential Equations," was accepted in IEEE International Conference on Big Data (IEEE BigData) 2023.
(October 28, 2023) One paper, titled "HyperNetwork Approximating Future Parameters for Time Series Forecasting under Temporal Drifts," was accepted in NeurIPS Workshop on Distribution Shifts (DistShift) 2023.
(October 20, 2023) One paper, titled "Continuous-time Autoencoders for Regular and Irregular Time Series Imputation," was accepted in ACM International Web Search and Data Mining Conference (WSDM) 2024 for its oral presentation.
(October 7, 2023) πThe paper of TableGAN, which is a pioneering work in tabular data synthesis, is now the most cited among all VLDB papers published within the last 5 years.
One paper, titled "Hypernetwork-based Meta-Learning for Low-Rank Physics-Informed Neural Networks," was accepted in Annual Conference on Neural Information Processing Systems (NeurIPS) 2023 for its spotlight presentation. [arXiv&GitHub]
One paper, titled "MadSGM: Multivariate Anomaly Detection with Score-based Generative Models," was accepted in the ACM International Conference on Information and Knowledge Management (CIKM) 2023.
One paper, titled "Precursor-of-Anomaly Detection for Irregular Time Series," was accepted in ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) 2023.
Prof. Noseong Park was invited to present GREAD, a recent ICML paper, on June 27th for SEA-CROGS (https://www.pnnl.gov/projects/sea-crogs) where Sandia National Labs and Pacific Northwest National Labs are collaborating to solve scientific problems with deep learning. [Webinar Schedule]
One paper, titled "Self-Attention LSTM-FCN Model for Arrhythmia Classification and Uncertainty Assessment," was accepted in Artificial Intelligence In Medicine (IF=7.011).Β
Two papers, titled "GREAD: Graph Neural Reaction-Diffusion Networks" and "CoDi: Co-evolving Contrastive Diffusion Models for Mixed-type Tabular Synthesis," were accepted in International Conference on Machine Learning (ICML) 2023. [arXiv&GitHub of GREAD] [arXiv&GitHub of CoDi]
One paper, titled "Hawkes Process based on Controlled Differential Equations," was accepted in International Joint Conference on Artificial Intelligence (IJCAI) 2023.
One paper, titled "Blurring-Sharpening Process Models for Collaborative Filtering," was accepted in International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) 2023. [arXiv&GitHub]
One paper, titled "Graph Neural Rough Differential Equations for Traffic Forecasting," was conditionally accepted in ACM Transactions on Intelligent Systems and Technology (IF=10.489).
One paper, titled "Fast Adversarial Training with Dynamic Batch-level Attack Control," was accepted in Design Automation Conference (DAC) 2023.
μ‘°λ―Όμ£Ό νμκ³Ό κ΅μΉμ§ νμμ΄ Hawkes Process μ°κ΅¬λ‘ μ 29ν μΌμ±ν΄λ¨Όν ν¬λ Όλ¬Έλμμμ CSE λΆμΌ μ₯λ €μμ μμνμ΅λλ€. μ°μΈλ μΈκ³΅μ§λ₯νκ³Όκ° "κ΅μλΉ μ΅λ€μμνκ³Ό (μ μκ΅μ λλΉ μμμ λΉμ¨)"λ‘ μ μ λμμ΅λλ€. [κ΄λ ¨ κΈ°μ¬]
One paper, titled "STaSy: Score-based Tabular Data Synthesis," was accepted for spotlight (notable-top-25%) presentation in International Conference on Learning Representations (ICLR) 2023. [arXiv&GitHub]
Seoyoung Hong won the best student paper award at IEEE BigData 2022 for her paper titled "TimeKit: A Time-series Forecasting-based Upgrade Kit for Collaborative Filtering." IEEE BigData ranked right after ICDM is one of the major conferences in the field of data mining & analysis. [κ΄λ ¨ κΈ°μ¬]
μ‘°λ―Όμ£Ό, μ§μμ°, μμ κ· νμμ΄ "2022λ λ κΈμ΅κ²°μ μλΉμ€ λ°μ λ Όλ¬Έ 곡λͺ¨μ (μ£Όμ΅: κΈμ΅κ²°μ μ)"μμ Neural CDEλ₯Ό νμ©ν κΈμ΅ μκ³μ΄ λ°μ΄ν° μμΈ‘ λ° μ΄μνμ§ μ£Όμ λ‘ μ΅μ°μμμ μμνμμ΅λλ€. [κ΄λ ¨ κΈ°μ¬]
λ°λ Έμ± κ΅μκ° μ λ 보μ₯ λΆκ΅μ(tenured associate professor) μΉμ§ νμμ΅λλ€.
One paper, titled "Learnable Path in Neural Controlled Differential Equations," was accepted for oral presentation in AAAI Conference on Artificial Intelligence (AAAI) 2023. [GitHub]
One paper, titled "TimeKit: A Time-series Forecasting-based Upgrade Kit for Collaborative Filtering," was accepted in IEEE International Conference on Big Data (IEEE BigData) 2022.
Yonsei AI μν¬μμμ BigDyL μ°κ΅¬μ€ νμλ€μ΄ μ¬λ¬ κ΅μλλ€κ³Ό μ°Έμ¬ κΈ°μ μ°κ΅¬μλ€μ μ¬μ¬μ μν΄ λ€μ μμ νμμ΅λλ€.
μ μ§μ± νμμ΄ GT-GAN: General Purpose Time Series Synthesis with Generative Adversarial Networks (NeurIPS 2022) λ Όλ¬ΈμΌλ‘ "LG AI μ°κ΅¬μ μ΅μ°μ MLμ°κ΅¬μ (λΆμ: iPad Air)"μ μμνμμ΅λλ€.
μ΄μ¬ν νμμ΄ LORD: Lower-Dimensional Embedding of Log-Signature in Neural Rough Differential Equations (ICLR 2022) λ Όλ¬ΈμΌλ‘ "λ€μ΄λ² Clova μ°μ MLμ°κ΅¬μ (λΆμ: ν΄λ‘λ° λ¨ν + 12κ°μ μ΄μ©κΆ)"μ μμνμμ΅λλ€.
μ΄μ¬ν νμμ΄ Invertible Tabular GANs: Killing Two Birds with One Stone for Tabular Data Synthesis (NeurIPS 2021) λ Όλ¬ΈμΌλ‘ "μΌμ±μ’ ν©κΈ°μ μ μ°μ MLμ°κ΅¬μ (λΆμ: κ°€λμ μμΉ5 νλ‘)"μ μμνμμ΅λλ€.
κΉμμ νμμ΄ SOS: Score-based Oversampling for Tabular Data (KDD 2022) λ Όλ¬ΈμΌλ‘ "LGμ μ μ°μ MLμ°κ΅¬μ (λΆμ: LG μ¨λ€λΉ)"μ μμνμμ΅λλ€.
μ‘°λ―Όμ£Ό νμκ³Ό μ΄μ¬ν νμμ΄ κ°κ° κ΄μ μ΄μ’ νκ΅μ‘μ¬λ¨κ³Ό SBSλ¬Ένμ¬λ¨μ μ₯νμμΌλ‘ μ λ°λμμ΅λλ€.
One paper, titled "GT-GAN: General Purpose Time Series Synthesis with Generative Adversarial Networks," was accepted in Conference on Neural Information Processing Systems (NeurIPS) 2022. [GitHub]
Two papers, titled "Attentive Neural Controlled Differential Equations for Time-series Classification and Forecasting" and "Climate Modeling with Neural Advection-Diffusion Equation," were accepted in Knowledge and Information Systems (IF=3.205). These papers are extensions of our last year's ICDM papers.Β
One paper, titled "Phishing URL Detection: A Network-based Approach Robust to Evasion," was accepted in the ACM Conference on Computer and Communications Security (CCS) 2022. [GitHub]
One full research paper, titled "Prediction-based One-shot Dynamic Parking Pricing," and one short research paper, titled "An Empirical Study on the Membership Inference Attack against Tabular Data Synthesis Models," were accepted in ACM International Conference on Information and Knowledge Management (CIKM) 2022.
One paper, titled "Two-Stage Training of Graph Neural Networks for Graph Classification," was accepted in Neural Processing Letters (IF=2.565).
One paper, titled "When Neural ODE Meets Adaptive Moment Estimation: Boosting Efficiency, Stability and Accuracy of Neural ODEs Together," was accepted in the ICML Workshop on Continuous Time Methods for Machine Learning 2022.
One paper, titled "SOS: Score-based Oversampling for Tabular Data," was accepted in ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) 2022. [GitHub]
One paper, titled "It's All In the Teacher: Zero-Shot Quantization Brought Closer to the Teacher," was accepted for oral presentation in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2022.
One paper, titled "SHIELD: Defending Textual Neural Networks against Black-Box Adversarial Attacks with Stochastic Multi-Expert Patcher," was accepted in Annual Meeting of the Association for Computational Linguistics (ACL) 2022.
μ°μΈλ λΉ λ μ°κ΅¬μ€μ λ°λ Έμ± κ΅μκ° μΌμ±ν΄λ¨Όν ν¬λ Όλ¬Έλμ μμ 곡λμ§λκ΅μμμ μ°μΈλ μ»΄ν¨ν°κ³Όνκ³Ό μ΄μ§νΈ, κΉμμ κ΅μλλ€κ³Ό ν¨κ» μμνμμ΅λλ€.
Our paper, titled "Graph Neural Controlled Differential Equations for Traffic Forecasting," was selected for oral presentation in AAAI 2022. The top 4.21% (380/9020) of papers were selected for oral presentation in AAAI 2022.
One paper, titled "Enabling Hard Constraints in Differentiable Neural Network and Accelerator Co-Exploration," was accepted in the 59th Design Automation Conference (DAC) 2022.
One paper, titled "LORD: Lower-Dimensional Embedding of Log-Signature in Neural Rough Differential Equations," was accepted in International Conference on Learning Representations (ICLR) 2022. We received a high average review score of 7.5, ranked 116/3328 (top 3%).
One paper, titled "Extrapolation and Interpolation-based Neural Controlled Differential Equations for Time-series Classification and Forecasting," was accepted in Web Conference (WWW) 2022. [GitHub]
Our two ICDM papers were invited for journal publications. In ICDM, a selected number of best papers are invited for possible inclusion, in an expanded and revised form, in the Knowledge and Information Systems journal (http://kais.bigke.org/) published by Springer, a top-tier journal in the field of data mining.
One paper, titled "Graph Neural Controlled Differential Equations for Traffic Forecasting," was accepted in AAAI Conference on Artificial Intelligence (AAAI) 2022. [Github]
νκ΅μΈκ³΅μ§λ₯νν/LG AIμ°κ΅¬μ μΆκ³κ³΅λνμ λνμ νν λ¦¬μΌ λ°ν λ° μ½λ μλ£μ λλ€. [Neural ODE] [Collaborative Filtering] [Next PoI Recommendation] [Neural CDE]
One paper, titled "Linear, or Non-Linear, That is the Question!," was accepted in ACM International Web Search and Data Mining Conference (WSDM) 2022. [Github]
Two papers, titled "Invertible Tabular GANs: Killing Two Birds with One Stone for Tabular Data Synthesis," and "Qimera: Data-free Quantization with Synthetic Boundary Supporting Samples," were accepted in Annual Conference on Neural Information Processing Systems (NeurIPS). IT-GAN received a high average review score of 7.0, ranked 380/9122 (top 4.1%). [GitHub of IT-GAN]
Prof. Noseong Park was awarded a research grant from the Okawa Foundation. [Link]
μ°μΈλ λΉ λ μ°κ΅¬μ€μ μ¬ ν΄ KDD, CIKM, ICDM λ Όλ¬Έλ€μ΄ Graph-based Recommendation, Next PoI Recommendataion, κ·Έλ¦¬κ³ λ§μ Time-series Classification & Forecasting λ°μ΄ν°μ μμ νμ¬ μ μΈκ³ λͺ¨λ κΈ°μ μ€ State-of-The-Art (SOTA) μ νλλ₯Ό 보μ΄κ³ μμ΅λλ€. λͺ¨λ κΉνμ΄ κ³΅κ°λμ΄ κ΄μ¬μλ νμλ€μ μ€μ€λ‘ νμ΅ κ°λ₯ν©λλ€. νκ΅μΈκ³΅μ§λ₯νν/LG AIμ°κ΅¬μ μΆκ³κ³΅λνμ λνμ νν 리μΌμ λ³Έ μ°κ΅¬μ€μμ μ§ννλλ°, κ΄λ ¨ν΄μ κ΅¬κΈ μ½λ©μμ μ½κ² μ€ν κ²°κ³Όλ€μ μ¬νν΄ λ³Ό μ μλ μ₯¬νΌν° λ ΈνΈλΆλ μ€λΉ μ€ μ λλ€.
Dr. Nur, who was advised by Prof. Noseong Park in the US, just has earned her Ph.D. and joined the Florida Institute of Technology as a tenure track assistant professor. Big congrats on her big success.
Two full papers, titled "Climate Modeling with Neural Diffusion Equations," and "Attentive Neural Controlled Differential Equations for Time-series Classification and Forecasting," were accepted in IEEE International Conference on Data Mining (ICDM) 2021. [Github of NDE] [Github of ANCDE]
μ°μΈλ λΉ λ μ°κ΅¬μ€μ λ°λ Έμ± κ΅μκ° ν¬ν곡과λν (POSTECH) μΈκ³΅μ§λ₯λνμμ κ²Έμ§κ΅μλ‘ λ°λ Ή λ°μμ΅λλ€. [λ§ν¬]
Jinsung Jeon just has finished his first year in our lab as a lab student representative. He has published 4 first-author and 3 co-author papers in top-tier conferences for the last only one year. Big congrats on his extraordinary performance!
Two full papers, titled "LT-OCF: Learnable-Time ODE-based Collaborative Filtering," and "LightMove: A Lightweight Next-POI Recommendation forTaxicab Rooftop Advertising," were accepted in ACM International Conference on Information and Knowledge Management (CIKM) 2021. [Github of LT-OCF] [Github of LightMove]
μ°μΈλ λΉ λ μ°κ΅¬μ€μ λ°λ Έμ± κ΅μκ° "2μ°¨ λ€μ΄λ² λ΄μ€ μκ³ λ¦¬μ¦ κ²ν μμν"μ μμμΌλ‘ μ°Έμ¬ν μμ μ λλ€. λ€μ΄λ²μ λ΄μ€ μΆμ² μκ³ λ¦¬μ¦μ λν 2018λ μ 1μ°¨ κ²ν μ μ΄μ 2μ°¨ κ²ν κ° μ¬ 8μ λΆν° μμλ μμ μ λλ€. [λ§ν¬]
Two papers, titled "ACE-NODE: Attentive Co-Evolving Neural Ordinary Differential Equations," and "Large-Scale Data-Driven Airline Market Influence Maximization," were accepted in ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) 2021. [Github of ACE-NODE]
A paper, titled "A Novel Method to Solve Neural Knapsack Problems," was accepted for its spotlight presentation by International Conference on Machine Learning (ICML) 2021. [Codes&Data in Bitbucket]
A paper, titled "A Sweet Rabbit Hole by DARCY: Using Honeypots to Detect Universal Trigger's Adversarial Attacks," was accepted in the Annual Meeting of Association for Computational Linguistics (ACL) 2021. [Github]
One project supported by Samsung and two projects supported by IITP had been granted.
Tabular data synthesis using GANs (Samsung, β©110M, 2021.04 ~ 2022.03)
Anomaly detection in time-series data (IITP, β©100M per year, 2021.04 ~ 2023.12)
Synopsis table synthesis using GANs, VAEs, invertible neural networks, and other generative models (IITP, β©120M per year, 2021.04 ~ 2024.12)
In addition, we also have several other funds supported by Yonsei University, BK21, Graduate School of AI, LG Display, Socar, and so on (In total, more than β©150M ~ 200M per year depending on situations).
Noseong Park will serve as a PC Co-Chair for the 20th Workshop on Privacy in the Electronic Society (WPES) which will be held in conjunction with the ACM Conference on Computer and Communications Security (CCS) 2021.
A paper, titled "OCT-GAN: Neural ODE-based Conditional Tabular GANs," was accepted in Web Conference (former WWW) 2021.
The Proceedings of the 16th EAI International Conference on Security and Privacy in Communication Networks (Securecomm) 2020 had been published. Noseong Park was the Publications Chair. Securecomm has CORE conference ranking: B, GII-GRIN-SCIE: B (Top 21%), LiveSHINE: A, H - Index: 42.
A paper, titled "Scalable Graph Synthesis with Adj and 1-Adj," was accepted in SIAM International Conference on Data Mining (SDM) 2021.
A paper, titled "Large-Scale Flight Frequency Optimization with Global Convergence in the US Domestic Air Passenger Markets," was accepted in SIAM International Conference on Data Mining (SDM) 2021.
A paper, titled "Attention-based Stress Detection exploiting Non-contact Monitoring of Movement Patterns with IR-UWB radar," was accepted in ACM/SIGAPP Symposium On Applied Computing (SAC) 2021.
A paper, titled "DPM: A Novel Training Method for Physics-Informed Neural Networks in Extrapolation," was accepted in AAAI Conference on Artificial Intelligence (AAAI) 2021. [Github]
A paper, titled "MONSTOR: An Inductive Approach for Estimating and Maximizing Influence over Unseen Social Networks," was accepted in IEEE/ACM International Conference on Advances in Social Network Analysis and Mining (ASONAM) 2020.