One-Hot Numerical Data Processing for Shorter Critical Paths
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional deep learning data analysis methods require complex hardware with increased circuitry area due to the need for more comparators, counters, and D-type Flip-Flop pipelines as input increases, leading to longer critical paths.
Innovation Solution
A method and device that transforms original numerical values into one-hot encoded number sequences, utilizing a shifter circuit, sorting circuit, and data distribution circuit to determine maximal, minimal, and survival values, reducing circuit complexity through parallel processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional hardware techniques are used to determine data distribution and sorting, then the data analysis function is achieved, but the circuitry area increases and critical paths become longer due to more comparators, counters, and D-type Flip-Flop pipelines
Solution Approach 1:
The patent transforms numerical data from its original format into one-hot encoding format, changing the parameter representation. This transformation allows the data to be processed using simpler logic circuits (AND gates and OR gates) instead of complex comparators and counters, thereby reducing circuitry area while maintaining data analysis capability
Solution Approach 2:
The patent replaces the mechanical/hardware-intensive system of comparators, counters, and D-type Flip-Flop pipelines with a logic-gate-based system using AND gates and OR gates. This substitution dramatically reduces the hardware complexity and critical path length while achieving the same data distribution and sorting functions
2Measurement precision
If more comparators and counters are used to handle increased input, then data distribution determination is more accurate, but the critical path length increases
Solution Approach 1:
By changing the data representation parameter to one-hot encoding, the patent enables parallel processing of multiple data points simultaneously through simple logic gates. This approach maintains measurement precision for data distribution while significantly reducing the critical path delay compared to sequential comparison methods
3Manufacturing precision
If a sorter network with higher complexity is required for sorting, then sorting accuracy is improved, but the hardware area increases
Solution Approach 1:
The patent changes the data format to one-hot encoding, which enables sorting to be achieved through simple logic operations (AND gates for intersection, OR gates for union) rather than complex comparator-based sorting networks. This maintains sorting accuracy while dramatically reducing hardware area requirements
Solution Approach 2:
The patent extracts the essential sorting function from complex comparator networks and implements it using minimal logic gates. By taking out only the necessary logical operations needed for sorting in one-hot encoding format, the hardware area is minimized while preserving sorting functionality
Data Source
AI summary
A numerical data processing method includes transforming a plurality of original numerical values of data into a plurality of number sequences; and determining a maximal value, a minimal value and at least a survival value of the plurality of number sequences according to a sorting result of the plurality of number sequences; wherein the plurality of number sequences are one-hot encoding.