pytc.df.pivots.pivoted_cholesky_streaming¶

pytc.df.pivots.pivoted_cholesky_streaming(diagonal, column_evaluator, shift, *, n_rank, batch_size=1, candidate_oversampling=1, n_topup=0)[source]¶

Select pivots from an exact diagonal and exact batched-column oracle.

candidate_oversampling chooses a stale-diagonal pool of candidate_oversampling * batch_size columns. Exact greedy re-pivoting retains only batch_size columns, followed by one blocked factor update. The final n_topup pivots use exact greedy singleton updates.