| 摘要 |
<p>A one dimensional control function is defined by generating points in error/cost space corresponding to parameter value combinations applied to reference input data and selecting a subset (2-11) in which each is such that all other points have a higher error or cost. Those combinations corresponding to the subset are associated with the values of a control variable so that a control variable value applied to the control function causes the respective combination of parameter values to be applied. The function can control a process (e.g. video standards converter) to optimise processing cost (e.g. time) and error. Also disclosed is controlling a process by the value of a control variable applied to a 1D control function which affects processing cost and error, where the variable values are associated in the function with selected sub-sets of control parameters for the process so that increasing values of the variable result in monotonically decreasing process errors and monotonically increasing cost. The incoming data may be segmented, and the local error/cost function for each segment may be estimated. The control may equalise the gradient of the function across all segments by changing the value of the control variable.</p> |
