Result Tables

Compact views of the generated experiment outputs. Downloadable CSVs are linked from the data page.

One-SE p-adic Ensemble Selections

dataset label	method	subset size s	ensemble size m	candidate library size K	mean padic loss	se padic loss	mean exact branch accuracy	mean active parameters used
Binary	greedy	9	4	200	0.0093	0.0026	0.8667	9.3917
Binary	prefix	7	94	200	0.0383	0.0024	0.6417	249.2
Multistate	greedy	4	4	200	0.0053	0.0025	0.9333	17.65
Multistate	prefix	5	194	200	0.046	0.0011	0.5167	1,022.1

Best Timing-Grid Rows by Loss

dataset label	method	subset size s	ensemble size m	mean padic loss	se padic loss	mean total fixed library elapsed seconds	mean estimated total k equals m elapsed seconds
Multistate	greedy	8	14	0.003	0.0024	7.3012	0.7285
Binary	greedy	9	5	0.0083	0.0029	1.4778	0.0889
Binary	prefix	7	94	0.0383	0.0024	6.3013	2.9656
Multistate	prefix	5	194	0.046	0.0011	9.3106	9.0318

Nested Branch-Point Models

dataset label	model family	model	outer folds	repeats	exact branch accuracy	mean padic loss	mean active parameters used
Binary	logistic_nested	nested_logistic_selected	12	1	0.5	0.0467	38.75
Binary	padic_greedy_nested	nested_padic_greedy	12	10	0.4167	0.2233	10.47
Multistate	logistic_nested	nested_logistic_selected	12	1	0.5	0.1133	91.25
Multistate	padic_greedy_nested	nested_padic_greedy	12	10	0.4417	0.1703	16.34

Logistic and Random-Forest Feature Comparison Models

dataset label	model group	outer folds	exact branch accuracy	mean padic loss	target absent folds
Binary	nested_logistic	12	0.5	0.0467	2
Binary	nested_random_forest	12	0.4167	0.05	2
Binary	nested_decision_tree	12	0.3333	0.16	2
Multistate	nested_random_forest	12	0.5	0.0467	2
Multistate	nested_logistic	12	0.5	0.1133	2
Multistate	nested_decision_tree	12	0.5	0.1267	2

All-Feature Tree Regression

target tree label	witnesses	features	target nodes	best model	best mean padic loss	best exact model	best exact mean padic loss	best logistic model	best logistic mean padic loss
All-label bridged all-feature tree	41	219	39	opaque_logistic_l1_C1	0.0572	padic_exact_s3_k100	0.1063	opaque_logistic_l1_C1	0.0572
Direct all-feature core tree	8	139	6	padic_exact_s2_k100	0.286	padic_exact_s2_k100	0.286	path_logistic_l2_C1	0.307

Within-Dataset Loss/Complexity Tradeoffs

dataset label	family points	slope log loss vs log params	r squared	linear p value	dummy loss	padic loss	best high capacity loss
Donker ath 1cor	5	-0.2487	0.8742	0.0197	0.2241	0.1602	0.0319
Donker ath 2cor titus	5	-0.2198	0.7213	0.0686	0.1662	0.1643	0.0219
Donker ath acts	5	-0.0898	0.8655	0.0218	0.1205	0.1111	0.0544
Donker ath acts 13 28	5	-0.1642	0.8406	0.0284	0.4437	0.3575	0.1133
Donker ath acts 1 12	5	-0.0744	0.8642	0.0222	0.0834	0.0669	0.0482
Donker ath catholicepistles	5	-0.2602	0.6203	0.1137	0.307	0.0527	0.0349
Donker ath heb	5	-0.2571	0.76	0.054	0.4073	0.354	0.0438
Donker ath paulineepistles	5	-0.2013	0.7657	0.052	0.3143	0.1856	0.0547
Donker ath rom	5	-0.1192	0.5604	0.1455	0.1751	0.1728	0.064

Donker p-adic Aggregation Tradeoffs

dataset label	family points	slope log loss vs log params	r squared	linear p value	spearman rho
Donker ath 1cor	5	-0.2123	0.8902	0.016	-0.9
Donker ath 2cor titus	5	-0.2029	0.8944	0.0151	-0.9
Donker ath acts	5	-0.0696	0.7103	0.0731	-0.9
Donker ath acts 13 28	5	-0.153	0.8981	0.0143	-0.7
Donker ath acts 1 12	5	-0.0734	0.8811	0.0181	-0.9
Donker ath catholicepistles	5	-0.2632	0.5489	0.1521	-0.7
Donker ath heb	5	-0.243	0.9183	0.0102	-1
Donker ath paulineepistles	5	-0.1993	0.8002	0.0405	-0.9
Donker ath rom	5	-0.1289	0.698	0.0781	-1

Donker p-adic Aggregation Outcome

dataset label	current padic loss	best padic policy	best padic loss	best euclidean group	best euclidean loss	best padic beats or ties euclidean
Donker ath 1cor	0.1602	no_intercept_deepest_member	0.0999	Random forest	0.0318	0
Donker ath 2cor titus	0.1643	no_intercept_branch_risk_deepest_tie	0.072	Random forest	0.0221	0
Donker ath acts	0.1111	no_intercept_branch_risk_deepest_tie	0.0767	Random forest	0.0544	0
Donker ath acts 13 28	0.3575	deepest_member	0.1608	Random forest	0.1133	0
Donker ath acts 1 12	0.0669	branch_risk_deepest_tie	0.0549	Logistic	0.0482	0
Donker ath catholicepistles	0.0527	branch_risk	0.0384	Random forest	0.0372	0
Donker ath heb	0.354	branch_risk_deepest_tie	0.0878	Random forest	0.0438	0
Donker ath paulineepistles	0.1856	branch_risk	0.1594	Random forest	0.0547	0
Donker ath rom	0.1728	no_intercept_branch_risk_deepest_tie	0.1687	Random forest	0.0639	0

External Corpus Loss/Complexity Tradeoffs

dataset label	panel id	family points	slope log loss vs log params	r squared	linear p value	dummy loss	padic loss	best high capacity loss
catena ephesians	panel_coverage	5	0.0697	0.0221	0.8114	0	0	0.0001
catena galatians	panel_coverage	5	-0.0002	0.0015	0.9509	0	0	0
catena romans	panel_coverage	5	0.0415	0.0076	0.8892	0.0007	0.0007	0.0013
comp tc mark stemma backbone	panel_coverage	5	-0.0442	0.1375	0.5389	0.3167	0.35	0.2333

External p-adic Aggregation Tradeoffs

dataset label	family points	slope log loss vs log params	r squared	linear p value	spearman rho
catena ephesians	5	0.0606	0.3082	0.3313	0.8947
catena galatians	5	0.0467	0.3778	0.27	0.8947
catena romans	5	0.0572	0.2893	0.3498	0.8947
comp tc mark stemma backbone	5	-0.0629	0.3298	0.3113	-0.5

External p-adic Aggregation Outcome

dataset label	current padic loss	best padic policy	best padic loss	best euclidean group	best euclidean loss	best padic beats or ties euclidean
catena ephesians	0	no_intercept_deepest_member	0	Random forest	0	1
catena galatians	0	no_intercept_deepest_member	0	Random forest	0	1
catena romans	0.0167	no_intercept_deepest_member	0.0167	Random forest	0.0167	1
comp tc mark stemma backbone	0.3167	drop_zero_members	0.2	Logistic	0.1505	0