About

Shengsheng Lin

Shengsheng Lin (林升升) is a Ph.D. student in Computer Science at South China University of Technology and is expected to graduate in June 2027. He is advised by Prof. Weiwei Lin, whose research focuses on cloud computing and big data.

His research lies at the intersection of time series forecasting and general-purpose AI. He focuses on efficient forecasting architectures, periodicity-aware modeling, and multivariate dependency learning, while also exploring large language models and multimodal methods for temporal data. His recent works include SparseTSF, CycleNet, SegRNN, and TQNet.

Research interests

  • Efficient architectures for time series forecasting
  • Periodicity-aware learning and temporal pattern analysis
  • Multivariate dependency learning for realistic forecasting
  • Large-scale foundation models for time series
  • Multimodal learning across time series, text, and vision

Selected Papers

Representative Publications

ICML 2024 Oral TPAMI 2026 2024 / 2026
GitHub stars for SparseTSF
Poster preview for SparseTSF

SparseTSF: Modeling Long-term Time Series Forecasting with 1k Parameters

Introduces cross-period sparse forecasting to reduce model parameters to as few as 1,000 while strengthening long-range dependency modeling, robustness, and generalization; now deployed in Didi ride-hailing demand-supply forecasting.

NeurIPS 2024 Spotlight 2024
GitHub stars for CycleNet
Poster preview for CycleNet

CycleNet: Enhancing Time Series Forecasting through Modeling Periodic Patterns

Explicitly models periodic patterns through a periodic-residual decomposition design, improving periodic utilization, residual modeling, and overall interpretability.

ICML 2025 2025
GitHub stars for TQNet
Poster preview for TQNet

Temporal Query Network for Efficient Multivariate Time Series Forecasting

Introduces temporal queries to capture global periodic patterns, enabling more robust multivariate dependency learning and stronger multivariate forecasting performance.

Education

Academic Training

South China University of Technology

Sep. 2022 - Present

Ph.D. Student in Computer Science

Conducting research on deep learning and artificial intelligence, with a particular focus on time series forecasting, efficient modeling, and periodic pattern analysis.

South China University of Technology

Sep. 2018 - Jun. 2022

B.E. Student in Computer Science

Built a systematic foundation in computer science, including computer systems, computer networks, software engineering, and core algorithmic training.

Experience

Industry and Engineering Practice

Didi

Dec. 2025 - Apr. 2026

Elite Program Algorithm Intern, Supply-Demand Scheduling Strategy Team

Optimized supply-demand forecasting models based on SparseTSF to address the high complexity of temporal dependency modeling in the original production system. Improved core business forecasting accuracy by 7%, reduced parameter count by 90%, and accelerated training by 5x, substantially speeding up model iteration and deployment.

Tencent

Jul. 2021 - Sep. 2021

Backend Development Intern, R&D Management Department

Developed backend services in Go using Gin, Docker, Nginx, Redis, Kafka, and XORM. Built and maintained scheduled-task interfaces and developed CLI tools with Cobra for unified task management.

Awards

Honors

  • National Scholarship (Ph.D. Student)
  • President Scholarship, South China University of Technology
  • National Scholarship (Undergraduate)
  • China International College Students' Innovation Competition Gold Award