SVDD Procedure

SOLVER Statement

SOLVER solver-type </options>;

The SOLVER statement specifies the type of optimization solver for SVDD training. You must specify one of the following solver-types; the default type is ACTSET.

ACTSET: uses the active-set solver for SVDD training. For more information about this solver, see the section Active-Set Solver.
STOCHS: uses a stochastic subset solver for SVDD training. For more information about this solver, see the section Stochastic Subset Solver.
FASTINC: uses a fast incremental solver for SVDD training. For more information about this solver, see the section Fast Incremental Solver.

You can specify the following option after a slash (/) for either solver-type:

MAXTIME=t: specifies an upper limit of t seconds for solving the SVDD optimization problem. If you omit this option, the solver does not stop based on the amount of time elapsed. The value of t can be any positive number; the default value is 1,800 seconds, or 30 minutes. This option puts a time budget only on the solvers, while the actual running time may be longer because of various kinds of overheads.

You can specify the following options after a slash (/) only when the solver-type is ACTSET or STOCHS:

MAXITER=number: specifies the maximum number of iterations of the stochastic solver, where number must be an integer greater than 0. To obtain a solution more quickly, specify a lower value for number. By default, MAXITER=25 for the active-set solver and MAXITER=200 for the stochastic subset solver.
STOL=number: specifies the tolerance value for the solver, where number must be a real number greater than 0. To obtain a solution more quickly, specify a higher value for number. By default, STOL=1E–4.

You can specify the following options after a slash (/) only when the solver-type is FASTINC or STOCHS:

MAXSV=number: specifies the maximum number of support vectors, where number must be an integer greater than 0. To obtain a solution more quickly, specify a lower value for number. By default, MAXSV=3500.

You can specify the following options after a slash (/) only when the solver-type is FASTINC:

OTHRESH=number

specifies the threshold for identifying outliers. The fast incremental solver considers an observation z to be an outlier if

max Underscript i Endscripts left-parenthesis upper K left-parenthesis x Subscript i Baseline comma z right-parenthesis right-parenthesis less-than sans-serif-italic number

where is the set of support vectors and

By default, OTHRESH=1E–20. The OTHRESH option is used only for the FASTINC solver and it is ignored for the ACTSET and the STOCHS solvers.

You can specify the following options after a slash (/) only when the solver-type is STOCHS:

CTOL=number: specifies the tolerance value that is used for detecting convergence of the center, where number must be a real number greater than 0. To obtain a solution more quickly, specify a higher value for number. By default, CTOL=1E–2.
NMATCH=number: specifies a convergence criterion for the stochastic solver, where number must be an integer greater than 1. If the radius and center values converge for the specified number of consecutive iterations, then convergence is declared. By default, NMATCH=5. To obtain a solution more quickly, specify a lower value for number.
NSAMP=number: specifies the number of observations that are sampled in each iteration of the stochastic solver, where number must be an integer greater than 0. To obtain a solution more quickly, specify a lower value for number. By default, NSAMP=10.
RTOL=number: specifies the tolerance value that is used to detect convergence of threshold value, where number must be a real number greater than 0. To obtain a solution more quickly, specify a higher value for number. By default, RTOL=1E–2.
SEED=number: specifies a seed value that is used for selecting a random sample, where number must be an integer. The default value of SEED is based on the system clock. If you want reproducible results, specify an integer for number.

Last updated: August 06, 2026