Matrix Compression with Adaptive Coefficients and Significant-Figure Rounding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data compression technologies face challenges in efficiently compressing large amounts of data while maintaining high classification accuracy and speed.
Innovation Solution
A data compression apparatus that generates a dictionary matrix, calculates compressed matrix data, and uses coefficients to round and correct elements based on significant figures, allowing for efficient data compression by determining coefficients based on the variation in non-zero elements and applying rounding operations accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing data compression technology is used, then data can be compressed, but compression efficiency is insufficient for large amounts of data
Solution Approach 1:
The patent segments the compression process into multiple stages: dictionary matrix generation from training data, compressed matrix calculation from original data, coefficient calculation based on statistical indices, and final rounding operations. This segmentation allows each stage to be optimized independently, improving overall compression efficiency for large datasets
Solution Approach 2:
The patent dynamically changes compression parameters based on data characteristics. Specifically, it calculates statistical indices (mean, standard deviation, kurtosis) from the compressed matrix and uses these to determine optimal rounding precision and coefficient values, allowing the compression algorithm to adapt to different data types and achieve higher efficiency
2Reliability
If compression is applied to maintain high classification accuracy, then classification accuracy is preserved, but compression speed decreases
Solution Approach 1:
The patent performs preliminary actions by pre-generating a dictionary matrix from training data before actual compression. This dictionary matrix contains pre-computed basis vectors that enable fast compression of new data. Additionally, statistical parameters (mean, standard deviation, kurtosis) are pre-calculated to guide subsequent rounding operations, ensuring both accuracy and speed
Solution Approach 2:
The patent extracts only the essential information needed for accurate classification by identifying and retaining non-zero elements in the compressed matrix. It calculates statistical indices from these extracted elements and uses them to determine minimum rounding precision, thereby maintaining classification accuracy while compressing data more efficiently and quickly
3Quantity of substance
If compression ratio is increased, then data size is reduced, but manufacturing precision of compressed data decreases
Solution Approach 1:
The patent dynamically adjusts compression precision parameters based on calculated statistical indices. It computes the mean, standard deviation, and kurtosis of the compressed matrix elements, then uses these to determine the optimal number of decimal places to retain. This adaptive parameter adjustment allows higher compression ratios while maintaining sufficient precision for the application
Solution Approach 2:
The patent applies different rounding precision levels to different elements of the compressed matrix based on their individual characteristics. Elements with higher statistical significance (determined by their contribution to reconstruction error or classification accuracy) retain more decimal places, while less critical elements are rounded more aggressively, achieving overall size reduction without sacrificing essential precision
Data Source
AI summary
A data compression apparatus includes a memory; and a processor configured to generate compressed matrix data, compare a threshold and an index value calculated about a specific value data string that is a data string obtained by coupling specific values specified from element values that are not zero values in each row of the compressed matrix data, specify a given constant as respective coefficients when the index value is larger than the threshold, calculate reciprocals of respective specific values as the respective coefficients when the index value is equal to or smaller than the threshold, and output matrix data after operation that is a result of rounding based on the number of places of significant figures of a decimal part in each element that corresponds about products of respective elements of the compressed matrix data and the respective coefficients calculated, regarding the respective elements of the compressed matrix data.