Complete result settings for Ad 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 5 -21.609221 -23.215033 -22.004436 27868 40081 6306 3661.897306
10 0.5 2 -18.410057 -20.015053 -18.934522 44772 80163 6388 3699.370609
10 0.5 1 -17.972938 -19.821236 -18.639057 57806 110353 6470 3688.970190
10 0.1 5 -25.301508 -28.625587 -26.616854 26875 37832 6322 3837.272645
5 0.5 5 -21.404080 -22.943431 -21.874061 62210 121065 6418 3809.062933
5 0.1 5 -24.471890 -27.728085 -25.734902 36934 61041 6488 4209.357079
2 0.5 5 -21.005752 -22.649901 -21.526531 77282 156241 6718 4406.977037
5 0.5 2 -17.384296 -18.976227 -18.052656 79784 160765 6662 4528.743525
10 0.1 1 -27.648512 -31.391478 -29.082627 46565 82775 6834 5233.244425
5 0.5 1 -16.809313 -18.526753 -17.624422 98230 205241 6876 5439.731034
10 0.5 0.1 -17.952258 -20.282347 -19.238646 117160 248394 7508 7817.209602
10 0.1 2 -26.907597 -30.564234 -28.505822 54069 98049 6862 5981.203718
10 0.1 0.1 -30.185337 -34.183885 -31.602638 67560 132080 7202 6254.343194
2 0.5 2 -15.867706 -17.313444 -16.656030 135444 289667 7504 10484.210157
5 0.1 2 -24.073253 -27.536037 -25.669897 78856 155754 7670 11110.902885
2 0.1 5 -22.702553 -25.699692 -24.197123 75352 147420 7334 11765.136426
5 0.5 0.1 -16.008218 -17.528600 -16.772854 155872 334092 8704 14872.081099
2 0.5 1 -14.817939 -16.392096 -15.626259 168034 367068 8124 15850.818309
2 0.5 0.1 -14.576074 -16.204972 -15.446457 253628 565827 9740 35554.808846
5 0.1 0.1 -26.386536 -31.033949 -28.977095 131088 271610 9276 37949.952730
5 0.1 1 -24.267697 -28.351900 -26.279472 127606 264572 9152 41746.910504
2 0.1 2 -20.665038 -24.106373 -22.645904 59016 102165 9766 75242.723367
2 0.1 1 -24.228516 -28.289638 -26.527875 50994 84478 9562 75416.909886
2 0.1 0.1 -23.480892 -27.856600 -25.803751 62408 106019 10996 75499.915267