Table 3.
Comparison of the running time (in seconds) for the QUIC, Big-QUIC, BCDIC and QUIC&Dirty methods on real-world data. The NNZ’s for the QUIC, Big-QUIC and BCDIC methods are around 10p by choosing their regularization parameter ρ.
| Data | p | ρ | NNZ | NLLte | QUIC | Big-QUIC | BCDIC | ρ1 | ρ2 | NLLte | QUIC&Dirty |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Snow | 9, 148 | 0.982 | 108,246 | 10639.1 | 282.3 | 243.5 | 72.0 | 0.982 | 10 | 10439.0 | 3174.2 |
| Air | 10, 512 | 0.975 | 105,636 | 12250.3 | 660.5 | 539.4 | 129.1 | 0.975 | 10 | 11965.7 | 7060.8 |
| Precip | 13, 610 | 0.820 | 132,200 | 13992.1 | 1680.9 | 1213.6 | 483.0 | 0.820 | − | − | − |
| OLR | 21, 720 | 0.988 | 223,936 | 25647.9 | 3650.2 | 2989.0 | 544.2 | 0.988 | − | − | − |
| Stock | 21, 602 | 0.985 | 221,053 | 24814.7 | 3990.6 | 3212.7 | 634.1 | 0.985 | − | − | − |
‘−’ indicates that the corresponding methods do not return a result after running over 5 hours for all choices of ρ2 ∈ {0.1, 1, 10}.