enhancementgood first issueimplementing algorithmsinterfacing algorithmsmodule:forecasting
仓库指标
- Star
- (7,162 star)
- PR 合并指标
- (平均合并 26天 10小时) (30 天内合并 86 个 PR)
描述
TabICL-TS is a hypothetical time series foundation obtained from applying make_reduction to tabICL, a tabular foundation model https://github.com/soda-inria/tabicl
This relates to TabPFN-TS as TabPFN relates to TabICL - we can expect higher accuracy at a given runtime budget.
It is insofar special as we would need to interface the tabular TabICl (in sklearn API) directly, and then delegate to a reducer composite.
Estimator interfacing guide: https://www.sktime.net/en/latest/developer_guide/add_estimators.html