Complete result settings for KDDCup 2000 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 0.1 0.1 -2.362270 -2.365572 -2.162380 295136 669755 1244 12297.995419
10 0.5 2 -2.357198 -2.362765 -2.157150 348689 788925 1320 15602.764018
5 0.1 2 -2.354616 -2.360061 -2.152333 369171 842753 1642 15906.486845
10 0.1 2 -2.363779 -2.366972 -2.162595 329674 742728 1196 17871.299148
10 0.1 1 -2.363005 -2.367589 -2.159780 347572 791636 1200 17243.061655
5 0.1 1 -2.352501 -2.358170 -2.153504 389477 888697 1688 19621.599062
10 0.5 1 -2.354874 -2.361307 -2.154915 394060 889106 1386 18208.421899
5 0.1 0.1 -2.351145 -2.356265 -2.153803 425531 967413 1760 19159.253349
10 0.5 0.1 -2.353489 -2.358983 -2.155778 458288 1031756 1432 21681.555901
5 0.5 2 -2.349658 -2.356184 -2.150962 430154 984729 1686 24073.806224
2 0.1 2 -2.345462 -2.351471 -2.148096 490373 1111103 2428 23718.707207
5 0.5 0.1 -2.344501 -2.350988 -2.148347 545331 1234044 1990 26146.042197
5 0.5 1 -2.345813 -2.352573 -2.149208 515926 1172390 1852 27695.817595
2 0.1 1 -2.343663 -2.349852 -2.146629 554248 1267612 2540 27633.574058
2 0.1 0.1 -2.339935 -2.345872 -2.142720 506550 1148131 2862 29483.551818
2 0.5 2 -2.339493 -2.347457 -2.141785 577898 1304843 2486 31123.690478
2 0.5 1 -2.336409 -2.345322 -2.138791 583671 1326156 2800 32805.136860
2 0.5 0.1 -2.332460 -2.343217 -2.142154 727400 1654405 3186 45981.714716
1 0.1 2 -2.339313 -2.347533 -2.141373 528050 1191886 3278 36990.996513
1 0.5 2 -2.334913 -2.344538 -2.138100 647658 1465588 3248 42392.852306
1 0.1 0.1 -2.333063 -2.345433 -2.138221 615454 1394208 4072 45269.156042
1 0.1 1 -2.336320 -2.346248 -2.139427 610730 1386984 3666 44373.010276
1 0.5 1 -2.331026 -2.342279 -2.139036 690723 1551790 3728 55513.743627