search

manish-pra / sampling-gpmpc

Sampling based GP MPC
MIT License
9 stars 0 forks source link

Avoid reinstantiating GPs for new data #4

Open lahramon opened 6 months ago

lahramon commented 6 months ago

Use the set_train_data method to update data points (fantasy_model seems not to work for multitask): https://docs.gpytorch.ai/en/stable/models.html#gpytorch.models.ExactGP.set_train_data

manish-pra commented 6 months ago

get_fantasy_model for multi-task GPs:

Resources: https://github.com/cornellius-gp/gpytorch/issues/800

tried in 2019 but failed some tests: https://github.com/cornellius-gp/gpytorch/pull/805

feature implemented in 2023: https://github.com/cornellius-gp/gpytorch/pull/2317