This is the base class for continuous scalar functions which needs cache for their pre-evaluated function returns.
The internal cache is created using the upper- and lower-bound domain values of the functional form (f(x))of subclasses of this class. So the cache only stores pre-evaluated values between f(lower-bound) and f(upper-bound).
To create a cache for continuous function, it uses sampling. With the given sample number , upper-bound, and lower-bound, it calculates interval within the ranges. It pre-evaluates and save f(x) where x = lower-bound + interval * [0 - sample number]
If a subclass of this class want to use a cache, it should explicitly call CreateCache(...) member function. GetCachedValue(x) will return pre-evaluated f(x) value. However, the return value from GetCachedValue(x) might be different from the exact return value from f(x) which is Evaluate(x) member function of subclasses of this class, because The GetCachedValue(x) member function internally converts x to cache table index and the conversion involves with truncation. So, users can think the cached value as an approximate to exact function return.
In some case, approximate values can be useful. For example, CompositeValleyFunction can be used as an M-estimator and it is currently used for MRIBiasFieldCorrectionFilter as an energy function. The bias field estimation requires calculation of energy values again and again for each iteration.
|typedef Array< MeasureType >||MeasureArrayType|
Public Member Functions
|virtual MeasureType||Evaluate (MeasureType x)|
|MeasureType||GetCachedValue (MeasureType x)|
Protected Member Functions
|void||CreateCache (double lowerBound, double upperBound, long sampleSize)|
|typedef double itk::CacheableScalarFunction::MeasureType|
Create the internal cache table and fill it with pre-evaluated values.
y = f(x) = (approximately) cache_table(index(x)) Get the function return using the internal cache table NOTE: Since the index calculation needs conversion from double to int, truncation happens. As a result, the return values from Evaluate(x) and GetCachedValue(x) may not be same for the same x.
Referenced by itk::CompositeValleyFunction::operator()().