## Problem Two `actg` features are functions of other features: - `str2` (antiretroviral history, 0 naive or 1 experienced) is exactly `strat > 1` for every patient. - `treat` (0 ZDV only, 1 other arms) is exactly `trt > 0` for every patient. A method can acquire the same information twice at twice the cost, and the feature count overstates what the dataset holds. ## Scope - Dataset generation for `actg`; both budget settings. - **Changes paper results** if features are dropped: `actg` is in the `kdd26` dataset list. ## Options to decide - Drop `str2` and `treat`, keeping the finer `strat` and `trt`. - Or keep all four as UCI ships them and state the redundancy on the dataset page. ## Acceptance criteria - `actg` follows the decision, and `docs/reference/datasets/actg.md` states it. - `just qa` passes.
Activity