[HIVEMALL-75] Support Sparse Vector Format as the input of RandomForest

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.5.0
• Assignee: Makoto Yui
• Reporter: Makoto Yui
• Created: 17/Feb/17 09:53
• Updated: 18/Jan/19 00:30
• Resolved: 30/Jun/17 12:16

Description

RandomForest implementation should support Sparse Vector as the input.

Matrix and CSRMatrix implementation is required for replacing double[][]x in
https://github.com/myui/incubator-hivemall/blob/master/core/src/main/java/hivemall/smile/classification/DecisionTree.java#L416

CSRMatrixBuilder should support incremental CSRMatrix construction.
https://github.com/myui/hivemall/commit/8f396178223809a8e8211070537c2539ebb6e165

Issue Links

• incorporates: HIVEMALL-233 RandomForest regressor accepts sparse vector input
• is blocked by: HIVEMALL-77 Support CSRMatrix and DenseMatrix
• relates to: HIVEMALL-128 Include version information is the prediction model built by RandomForest to support prediction for each version
• relates to: HIVEMALL-127 Create tree_predict_v1 for RandomForest model backward compatibility
• links to: GitHub Pull Request #51
• links to: GitHub Pull Request #70