Samples a relation matrix.
This widget samples the input data and sends both the sampled and the remaining data to the output. It is useful for evaluating the performance of recommendation systems.
- Select the desired sampling method:
- rows (randomly samples entire matrix rows)
- columns (randomly samples entire matrix columns)
- rows and columns (samples from the entire matrix)
- entries (randomly samples individual matrix elements)
- Select the proportion of the data you want at the output.
- Press Apply to commit the changes.
Matrix Sampler widget samples data into two subsets: in-sample and out-of-the-sample data. This is useful if you want to check the accuracy of matrix reconstruction with Completion Scoring. Feed in-sample data into the Fusion Graph to reconstruct the matrix and then compare the results with out-of-the-sample data.