mlrMBO contains most of the most popular infill criteria, e.g., expected
improvement, (lower) confidence bound etc. Moreover, custom infill criteria
may be generated with the makeMBOInfillCrit
function.
makeMBOInfillCritMeanResponse()
makeMBOInfillCritStandardError()
makeMBOInfillCritEI(se.threshold = 0.000001)
makeMBOInfillCritCB(cb.lambda = NULL)
makeMBOInfillCritAEI(aei.use.nugget = FALSE, se.threshold = 0.000001)
makeMBOInfillCritEQI(eqi.beta = 0.75, se.threshold = 0.000001)
makeMBOInfillCritDIB(cb.lambda = 1, sms.eps = NULL)
makeMBOInfillCritAdaCB(cb.lambda.start = NULL, cb.lambda.end = NULL)
[numeric(1)
]
In order to avoid numerical problems the standard error estimation is assumed to
be exactly zero, if it is below se.threshold
.
Default is 1e-6.
[numeric(1)
| NULL
]
Lambda parameter for confidence bound infill criterion.
Default is NULL
, which means 1 in case of a fully numeric parameter set and 2 otherwise.
Any non-negative real valued numbers are valid.
[logical(1)
]
Should the nugget effect be used for the pure variance estimation for augmented
expected improvement?
Default is FALSE
.
[numeric(1)
]
Beta parameter for expected quantile improvement criterion.
Default is 0.75.
Valid values are between 0.5 and 1.
[numeric(1)
| NULL
]
Epsilon for epsilon-dominance for dib.indicator = "sms"
.
Default is NULL
, in this case it is adaptively set.
[numeric(1)
| NULL
]
The value of cb.lambda
at the beginning of the optimization.
The makeMBOInfillCritAdaCB
crit takes the progress of the optimization determined by the termination criterion to linearly move from cb.lambda.start
to cb.lambda.end
.
The initial desgin does not account for the progress of the optimization.
Eexcept for makeMBOTerminationMaxExecBudget
) if you dont pass a precalculated initial design.
[numeric(1)
| NULL
]
The value of cb.lambda
at the end of the optimization.
In the multi-objective case we recommend to set cb.lambda
to
\(q(0.5 \cdot \pi_{CB}^{(1 / n)})\) where \(q\) is the quantile
function of the standard normal distribution, \(\pi_CB\) is the probability
of improvement value and \(n\) is the number of objectives of the considered problem.