Reproduce “Time-FFM: Towards LM-Empowered Federated Foundation Model for Time Series Forecasting” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Qingxiang Liu, Xu Liu, Chenghao Liu, Qingsong Wen, Yuxuan Liang
Time-FFM forecasts across domains by mapping time-series patches and selected embedding prototypes into a frozen language model. Federated clients share adaptation parameters while keeping prediction heads local and raw records on-device. In zero-shot tests, its mean MSE was 0.310 across ETTh2, Electricity, and Weather, versus FedIT’s 0.360; source-model selection used target-domain comparisons, and parameters are communicated.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.