Decision Tree Inference Parallel Processing via Data Row Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing decision tree inference processes are not efficiently parallelized, leading to suboptimal processing speeds despite the use of SIMD methods.
Innovation Solution
An information processing apparatus and method that divides input data rows based on condition determination results, grouping rows by feature amounts to enable parallel processing at condition determination nodes, thereby speeding up the inference process by performing comparisons and predictions in parallel.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If all condition determination nodes are processed using all sets of data, then the decision tree inference can be completed, but the processing efficiency is low
Solution Approach 1:
The patent segments the data processing by dividing the input data into multiple subsets and assigning different condition determination nodes to process different subsets. This segmentation enables parallel processing where each node handles a specific portion of the data, significantly improving processing efficiency and reducing inference time compared to processing all data at every node sequentially.
Solution Approach 2:
The patent introduces a new dimension to the traditional decision tree structure by adding a data subset dimension. Instead of processing all data rows at every node, the system processes different data subsets at different nodes simultaneously, creating a multi-dimensional processing approach that enables parallel computation and improves overall productivity.
2Speed
If parallel processing is applied to decision tree inference, then processing speed increases, but the system complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the data into subsets and assigning them to different nodes, which simplifies the parallel processing architecture. Each node handles a specific data subset, making the system more manageable and easier to implement compared to complex global parallel processing approaches, while still achieving speed improvements.
Solution Approach 2:
The patent makes the condition determination nodes universal by enabling them to handle both data filtering and parallel processing functions. The same node structure that traditionally only filtered data now also manages data subset assignment and parallel execution, reducing the need for separate complexity-managing components and simplifying the overall system.
Data Source
AI summary
In the information processing apparatus, an acquisition unit acquires an input data matrix including a plurality of data rows each including a plurality of feature amounts. A division unit generates grouping information by dividing at least a portion of row numbers of the input data matrix in association with a child node selected based on a result of a condition determination at a condition determination node, and passes the grouping information to the child node. A rearrangement process unit performs a condition determination process of the plurality of data rows indicated in the received grouping information by a parallel process at the condition determination node. An output unit outputs predicted values corresponding to the plurality of data rows indicated in the received grouping information at the leaf node.