
Parallel Colt 0.7.2  
PREV PACKAGE NEXT PACKAGE  FRAMES NO FRAMES 
See:
Description
Interface Summary  

FloatBinBinFunction1D  Interface that represents a function object: a function that takes two bins as arguments and returns a single value. 
FloatBinFunction1D  Interface that represents a function object: a function that takes two bins as arguments and returns a single value. 
Class Summary  

AbstractFloatBin  Abstract base class for all arbitrarydimensional bins consumes float elements. 
AbstractFloatBin1D  Abstract base class for all 1dimensional bins consumes float elements. 
DynamicFloatBin1D  1dimensional rebinnable bin holding float elements; Efficiently computes advanced statistics of data sequences. 
FloatBinFunctions1D  Function objects computing dynamic bin aggregations; to be passed to generic methods. 
MightyStaticFloatBin1D  Static and the same as its superclass, except that it can do more: Additionally computes moments of arbitrary integer order, harmonic mean, geometric mean, etc. 
QuantileFloatBin1D  1dimensional nonrebinnable bin holding float elements with scalable quantile operations defined upon; Using little main memory, quickly computes approximate quantiles over very large data sequences with and even without apriori knowledge of the number of elements to be filled; Conceptually a strongly lossily compressed multiset (or bag); Guarantees to respect the worst case approximation error specified upon instance construction. 
StaticFloatBin1D  1dimensional nonrebinnable bin consuming float elements; Efficiently computes basic statistics of data sequences. 
Multisets (bags) with efficient statistics operations defined upon; This package requires the Colt distribution.
Bins contain information about the data filled into them. They can be asked for various descriptive statistical measures, such as the minimum, maximum, size, mean, rms, variance, etc.
Bins come in two flavours: Dynamic and Static. Dynamic bins preserve
all the values filled into them and can return these exact values, when asked
to do so. They are rebinnable.
Static bins do not preserve the values filled into them. They merely collect
basic statistics incrementally while they are being filled. They immediately
forget about the filled values and keep only the derived statistics. They are
not rebinnable.
The data filled into static bins is not preserved. As a consequence infinitely
many elements can be added to such bins. As a further consequence such bins
cannot compute more than basic statistics. They are also not rebinnable. If
these drawbacks matter, consider to use a DynamicFloatBin1D
,
which overcomes them at the expense of increased memory requirements.
The data filled into dynamic bins is fully preserved.
Technically speaking, they are kdimensional multisets (or bags) with efficient statistics operations defined upon.
As a consequence such bins can compute more than only basic statistics.
They are also rebinnable.
On the other hand side, if many elements are filled into them, one may quickly run out of memory (each float element takes 4 bytes).
If these drawbacks matter, consider to use a StaticFloatBin1D
,
which overcomes them at the expense of limited functionality.
AbstractFloatBin
.
This base class is extended by AbstractFloatBin1D
, the common
abstract base class for 1dimensional bins.
Static 1dimensional bins currently offered are: StaticFloatBin1D
,
MightyStaticFloatBin1D
and QuantileFloatBin1D
.
Dynamic 1dimensional bins currently offered are: DynamicFloatBin1D
.
In case not each and every statistics measure needed is directly provided by
methods of bins one can use dynamic bins and retrieve their filled elements.
From these elements, one can compute whatever necessary, either by using a statistics
library or self written functions. Use methods like DynamicFloatBin1D.elements()
and, for example, the descriptive statistics library FloatDescriptive
.

Parallel Colt 0.7.2  
PREV PACKAGE NEXT PACKAGE  FRAMES NO FRAMES 