Sparse inverse-problem marathon with covariate shift on the underlying linear operator. Recover a 200-dim…
Sparse inverse-problem marathon with covariate shift on the underlying linear operator. Recover a 200-dim…: a task in Long-Horizon Terminal-Bench (LHTB) (Harbor dataset). You are given a synthetic dataset where each row contains 50 numeric "observations" y 0, ..., y 49 plus 200 numeric "signals" x…
The task
You are given a synthetic dataset where each row contains 50 numeric "observations" `y_0, ..., y_49` plus 200 numeric "signals" `x_0, ..., x_199`. The signals are sparse — most entries are zero — and the observations are produced by some hidden linear forward operator applied to the signals plus noise. Your job is to: