Sparse Data Structure Compression with Mode Search Approximation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional delta compression methods fail to provide satisfactory results for sparse data structures like sparse matrices used in machine learning due to large differences between neighboring numerical values, defeating traditional compression schemes.

Innovation Solution

A data compression method utilizing mode search approximation, where the most repeated value (mode) is extracted and encoded with a significance map, complementing delta compression, and employing a ternary tree processing algorithm for rapid mode approximation with parallel machine processing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If conventional delta compression is applied to sparse data structures with large differences between neighboring values, then compression is attempted, but compression efficiency deteriorates significantly

Engineering Contradiction:
Improvedata sizeVSAvoidcompression efficiency
Core Design Contradiction:
Quantity of substanceVSProductivity

Solution Approach 1:

The patent changes the compression parameter from delta encoding to mode-based encoding. Instead of storing differences between consecutive elements, the system identifies the mode (most frequent value) in data blocks and encodes deviations from this mode, which is particularly effective for sparse data structures where many elements share the same value.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent divides the data into blocks and processes each block independently to find the local mode. This segmentation allows the compression algorithm to adapt to local data characteristics and handle sparse structures more effectively by identifying repeated values within each block.

Inventive Principle:
Principle #1Segmentation

2Productivity

If mode search approximation is implemented to extract and encode the most repeated value, then compression efficiency for sparse data structures improves, but computational complexity increases

Engineering Contradiction:
Improvecompression efficiencyVSAvoidcomputational complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent performs mode search only on a subset of data blocks rather than the entire dataset. By selectively applying mode-based compression to blocks that benefit most from this approach, the system achieves good compression ratios while limiting computational overhead.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The patent pre-processes data to identify candidate mode values before performing the full compression encoding. This preliminary identification step simplifies the subsequent encoding process and reduces the overall computational complexity by avoiding repeated full scans of the data.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3719669B1Compression for sparse data structures utilizing mode search approximation
Publication Date: 2025.10.29 INTEL CORP
  • EP3719669B1 patent drawingFigure 1
  • EP3719669B1 patent drawingFigure 2
  • EP3719669B1 patent drawingFigure 3

AI summary

Embodiments are generally directed to compression for compression for sparse data structures utilizing mode search approximation. An embodiment of an apparatus includes one or more processors including a graphics processor to process data; and a memory for storage of data, including compressed data. The one or more processors are to provide for compression of a data structure, including identification of a mode in the data structure, the data structure including a plurality of values and the mode being a most repeated value in a data structure, wherein identification of the mode includes application of a mode approximation operation, and encoding of an output vector to include the identified mode, a significance map to indicate locations at which the mode is present in the data structure, and remaining uncompressed data from the data structure.