Ensemble Evaluation Cores

Once the ensemble is buit and combination strategy is chosen it can start classifying unseen instances. The instance classification process can be divided into two steps:

• first, every ensemble member must make a classificaiton of the current instances, this process is often called "ensemble member evaluation" process,
• second, classifications of individual ensemble members must be combined together make a final classification

There are two approaches for the realisation of ensemble:

• parallel, where every member from the ensemble is implemented as separate module and
• sequential, where one module implements all members of the ensemble.

So-Logic offers IP cores that can be used to implement ensemble member evaluation process directly in hardware, in case when ensemble is composed from decision trees. So-Logic offers IP cores for parallel and sequential ensemble member evaluation. Below you can find a list of currently available ensemble member evaluation cores:

• Decision Tree Ensemble Evaluation Core - Sequential Evaluation using Pipelined Architecture
• Decision Tree Ensemble Evaluation Core - Parallel Evaluation using Pipelined Architecture
• Decision Tree Ensemble Evaluation Core - Sequential Evaluation using Serial Architecture
• Decision Tree Ensemble Evaluation Core - Parallel Evaluation using Serial Architecture