Library workflow - 5.2 English - 68552

The intended workflow for using the AOCL-DA C APIs is as follows:

1. Load data from memory. Data can be obtained from various sources:

• pass arrays of data directly from another part of your application;

• read directly from CSV files into floating-point or integer arrays;

• read mixed data from CSV files into a da_datastore.

2. Data preprocessing. Functions are available for:

• standardizing, scaling and shifting arrays of floating point data;

• removing missing data from a da_datastore;

• selecting certain subsets of the data in a da_datastore;

• extracting contiguous arrays of data from a da_datastore.

3. Data processing. This is often split into several function calls:

• initialize a da_handle struct, which is used internally to store algorithmic information;

• pass arrays of data, or data extracted from a da_datastore, to the handle (for best possible performance, algorithmic functions typically operate on two-dimensional arrays stored in column major format);

• computation (e.g. clustering, linear model, principal component analysis);

• extract results from the da_handle.