Table 3.

Fractional Factorial Design Benchmark. Order and context of tuned parameters to optimize. the prediction quality in explainable and recoverable manner [Abbreviations: LR = Learning Rate; ML = Machine Learning].

Factor	Actions
nepoch = 1000	Small initial LR (<default) test all ML optimization algorithms fixed: LR, fModelSpread, nModelDepth
LR = 10−6..10−1 (exp. step size)	Increase LR step-wisely test all ML optimization algorithms with different LR fixed: nepoch, fModelSpread, nModelDepth
fModelSpread = 1..6 nModelDepth = 1..6	Increase fModelSpread and fModelSpread step-wisely test all ML optimization algorithms for different model sizes fixed: nepoch, LR
t alg	Test all ML optimization algorithms fixed: nepoch, LR, fModelSpread, nModelDepth