High Dynamic Ratio Field Compression With Adaptive Data Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing compression hardware and software are inflexible and inefficient for handling data with high dynamic ratio fields, particularly in machine learning applications, as they are designed for specific quantum sizes and struggle with larger data formats like 16-bit and 32-bit floating point data.
Innovation Solution
The method involves dynamically mapping M-bit data units into S data fields, using machine learning algorithms and directed search engines to select optimal mappings based on locality and compression ratios, allowing for flexible compression and decompression of data fields using different compression methods, and aggregating compressed data into memory transactions for efficient storage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If existing compression hardware and software are used with fixed quantum sizes, then compression efficiency is maintained for standard data formats, but flexibility and capability to handle larger data formats (16-bit and 32-bit floating point) are lost
Solution Approach 1:
The patent implements dynamic quantum size selection where the compression system can adaptively choose different quantum sizes (8-bit, 16-bit, 32-bit, or 64-bit) based on the input data characteristics. This allows the compression hardware to handle various data formats flexibly while maintaining optimization for each format, resolving the contradiction between adaptability and device complexity.
Solution Approach 2:
The system changes the quantum size parameter dynamically according to the data being compressed. By allowing the quantum size to vary from fixed 8-bit to larger values (16-bit, 32-bit, 64-bit), the compression hardware can efficiently handle different data formats without requiring completely different hardware configurations, thus improving adaptability while controlling complexity.
2Productivity
If data is divided into multiple data fields for compression, then compression ratio is improved for high dynamic ratio fields, but mapping complexity and processing overhead increase
Solution Approach 1:
The patent segments the bit stream of each data unit into multiple data fields (first data field, second data field, etc.) based on the mapping. This segmentation allows different compression methods to be applied to different fields, improving the overall compression ratio for high dynamic ratio fields while managing mapping complexity through systematic organization.
Solution Approach 2:
The system applies different compression methods to different data fields based on their specific characteristics. By analyzing the local properties of each data field and applying appropriate compression techniques selectively, the system improves compression ratios while avoiding unnecessary complexity in the mapping process.
3Productivity
If machine learning algorithms are used to dynamically generate mappings, then compression optimization is improved, but computational overhead and processing time increase
Solution Approach 1:
The patent uses machine learning algorithms to pre-generate and store optimal mappings for different data patterns and compression scenarios. By performing the computationally intensive mapping optimization in advance and caching the results, the system achieves compression optimization without incurring excessive processing time during actual compression operations.
Solution Approach 2:
The system employs machine learning with feedback mechanisms that analyze compression results and continuously optimize mapping strategies. By using feedback from actual compression performance, the system learns to generate better mappings over time, improving compression optimization while reducing the computational overhead through learned patterns.
Data Source
AI summary
Various embodiments include methods and devices for implementing decompression of compressed high dynamic ratio fields. Various embodiments may include receiving compressed first and second sets of data fields, decompressing the first and second compressed sets of data fields to generate first and second decompressed sets of data fields, receiving a mapping for mapping the first and second decompressed sets of data fields to a set of data units, aggregating the first and second decompressed sets of data fields using the mapping to generate a compression block comprising the set of data units.