Developer Guide for Intel® Data Analytics Acceleration Library 2016 Update 4
An algorithm can accept one or several numeric tables or models as an input. In computation modes that permit multiple calls to the compute() method, ensure that the structure of the input data, that is, the number of features, their order, and type, is the same for all the calls. The following methods are available to provide input to an algorithm:
input.set(Input ID, InputData) |
Use to sets a pointer to the input argument with the Input ID identifier. This method overwrites previous input pointer stored in the algorithm. |
input.add(Input ID, InputData) |
Use in the distributed computation mode to add the pointers with the Input ID identifier. Unlike the input.set() method, input.add() does not overwrite the previously set input pointers, but stores all the input pointers until a call to the compute() method. |
input.get(Input ID) |
Use to get a reference to the pointer to the input data with the Input ID identifier. |
For the input that each specific algorithm accepts, refer to the description of this algorithm.