Complete result settings for MSNBC dataset (table columns are sortable)
The experiments are described in Rooshenas and Lowd, Discriminative Structure Learning of Arithmetic Circuits, AIStats 16

Split Penalty Standard Deviation L1 Penalty Train LL Validation LL Test LL Node# Edge# Feature# Learning Time
10 0.5 1 -6.051743 -6.059390 -6.060397 179215 379751 3896 10775.309903
5 0.1 0.1 -6.040918 -6.048484 -6.051536 218643 458941 6532 14038.572811
5 0.5 1 -6.042201 -6.049471 -6.053367 216028 453584 6016 13397.073334
10 0.1 0.1 -6.058628 -6.064711 -6.066737 189036 398324 4440 12461.053630
10 0.5 2 -6.058770 -6.065021 -6.066381 225385 477121 4860 14737.954489
5 0.1 1 -6.042949 -6.050579 -6.053729 213136 447433 6068 14230.757595
5 0.5 0.1 -6.039237 -6.047303 -6.050448 220884 461567 6454 14162.223014
10 0.5 0.1 -6.056835 -6.062518 -6.065749 192903 413407 3874 13848.963636
5 0.1 2 -6.044666 -6.052292 -6.055426 231432 482256 7096 15933.978666
10 0.1 1 -6.054268 -6.060790 -6.062258 230948 490642 4522 14328.233776
10 0.1 2 -6.056106 -6.062633 -6.065242 219961 466810 4916 16010.646011
5 0.5 2 -6.042962 -6.050217 -6.053424 238496 497814 6702 15787.255972
2 0.1 2 -6.032157 -6.042169 -6.044539 259132 536131 8978 18403.150295
2 0.5 0.1 -6.029674 -6.039660 -6.043728 213127 431453 10034 21393.893633
2 0.5 1 -6.029443 -6.041624 -6.042724 256569 533222 8766 21150.227675
2 0.1 1 -6.031088 -6.040028 -6.043549 246308 511440 8502 20232.547184
2 0.5 2 -6.030522 -6.041316 -6.043351 269632 562672 8648 22181.221941
2 0.1 0.1 -6.030564 -6.041516 -6.043746 257892 537260 8772 25753.488874
1 0.1 0.1 -6.046352 -6.053788 -6.057332 136442 265893 10034 9928.474642
1 0.5 0.1 -6.057983 -6.063256 -6.067297 141611 276507 10034 11161.465198
1 0.1 2 -6.043408 -6.051476 -6.054212 148601 292828 10034 13056.885051
1 0.1 1 -6.040295 -6.049520 -6.050925 149219 293596 10034 13369.484529
1 0.5 2 -6.041608 -6.050449 -6.052443 143861 282100 10034 13351.934196
1 0.5 1 -6.041521 -6.050247 -6.052735 145667 285676 10034 13800.876947