CNN Processing Circuit Valid Pair Detection Mask Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep neural networks, particularly convolutional neural networks (CNNs), face high computational complexity, leading to increased power consumption and operation time.
Innovation Solution
A processing circuit and method that efficiently detect valid pairs between two inputs in a neural network operation by using a processing element (PE) with input circuits providing compressed chunks and masks, allowing for reduced computational load through selective operation on valid value pairs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional CNN computation is performed on all input values, then complete feature detection is achieved, but power consumption and computational complexity increase significantly
Solution Approach 1:
The patent extracts and processes only valid value pairs from the input chunks using mask circuits. By identifying and isolating the subset of input values that actually contribute to the output (valid pairs), the system avoids computing with invalid zeros, thereby reducing power consumption while maintaining feature detection accuracy.
Solution Approach 2:
The patent applies partial action by performing computations only on the necessary portion of input data (valid pairs) rather than all input values. The mask-based filtering enables the system to execute computations selectively, reducing overall computational load and power consumption without compromising the completeness of feature detection.
2Reliability
If conventional CNN computation is performed on all input values, then complete feature detection is achieved, but operation time increases
Solution Approach 1:
The patent extracts and processes only valid value pairs from the input chunks using mask circuits. By identifying and isolating the subset of input values that actually contribute to the output (valid pairs), the system avoids computing with invalid zeros, thereby reducing operation time while maintaining feature detection accuracy.
Solution Approach 2:
The patent applies partial action by performing computations only on the necessary portion of input data (valid pairs) rather than all input values. The mask-based filtering enables the system to execute computations selectively, reducing overall operation time without compromising the completeness of feature detection.
3Productivity
If compressed chunks with only valid values are processed, then computational load is reduced, but data structure complexity increases
Solution Approach 1:
The patent introduces mask circuits as intermediary components that facilitate the processing of compressed chunks. These mask circuits act as mediators between the compressed input data and the processing elements, enabling efficient identification and processing of valid value pairs while managing the complexity of the compressed data structure.
Solution Approach 2:
The patent segments the input data into compressed chunks with associated masks, dividing the processing task into manageable units. Each chunk-mask pair can be processed independently by processing elements, which simplifies the overall computation while handling the complexity of compressed data through structured segmentation.
4Use of energy by moving object
If valid pair detection is performed using sub-masks, then power consumption is reduced, but circuit complexity increases
Solution Approach 1:
The patent segments the mask into multiple sub-masks that correspond to different portions of the input data. This segmentation allows processing elements to efficiently detect valid pairs by comparing only relevant sub-masks, reducing power consumption through selective processing while managing circuit complexity through structured organization of mask components.
Solution Approach 2:
The patent applies partial action by using only the necessary sub-masks for valid pair detection rather than processing the entire mask at once. This selective approach reduces power consumption by activating only the circuit components needed for current computations, while the modular sub-mask structure helps manage overall circuit complexity.
Data Source
AI summary
Provided is an operating method of a processing circuit, the method including generating a first compressed chunk including only a first valid value, generating a first mask includes a reference value at a same position as a position of the first valid value, and includes a plurality of first sub-masks, generating a second compressed chunk including only a second valid value, generating a second mask, includes a reference value at a same position as a position of the second valid value, and includes a plurality of second sub-masks, generating a valid pair position value for each of a current first sub-mask and a current second sub-mask, generating a first cumulative value corresponding to a number of reference values included in a first previous sub-mask, and generating a second cumulative value corresponding to a number of reference values included in a second previous sub-mask.