18345533. METHOD AND APPARATUS WITH WEIGHT COMPRESSION simplified abstract (SAMSUNG ELECTRONICS CO., LTD.)
Contents
METHOD AND APPARATUS WITH WEIGHT COMPRESSION
Organization Name
Inventor(s)
METHOD AND APPARATUS WITH WEIGHT COMPRESSION - A simplified explanation of the abstract
This abstract first appeared for US patent application 18345533 titled 'METHOD AND APPARATUS WITH WEIGHT COMPRESSION
Simplified Explanation
The patent application describes a method and apparatus for optimizing the performance of filters in a system. Here are the key points:
- The method involves rearranging a set of filters and compressing their weights based on the result of the rearrangement.
- The compression process results in some weights remaining uncompressed.
- The uncompressed weights are then mapped to specific operation units based on a predetermined bulk unit, generating a set of operation unit maps.
- Finally, these operation unit maps are mapped to an array.
Potential applications of this technology:
- This technology can be applied in various fields where filters are used, such as image processing, signal processing, and machine learning.
- It can improve the efficiency and performance of systems that rely on filters, leading to faster and more accurate results.
Problems solved by this technology:
- The method addresses the issue of inefficient filter arrangements by reordering them, which can improve the overall performance of the system.
- By compressing weights, the method reduces the memory footprint required for storing the filters, making the system more efficient.
Benefits of this technology:
- The rearrangement and compression of filters can lead to faster processing times and improved accuracy in tasks that rely on filters.
- The reduced memory footprint allows for more efficient storage and utilization of filters, resulting in cost savings and improved system performance.
Original Abstract Submitted
A method and apparatus are provided. The method includes reordering a plurality of filters, then based on a result of the reordering, compressing weights, among a plurality of weights of the plurality of filters, resulting in some of the plurality of weights being uncompressed weights, generating a plurality of operation unit maps by mapping the uncompressed weights to respective operation units according to a predetermined bulk unit, and mapping the plurality of operation unit maps to an array.