18397716. APPARATUS AND METHOD OF PERFORMING MATRIX MULTIPLICATION OPERATION OF NEURAL NETWORK simplified abstract (Samsung Electronics Co., Ltd.)
Contents
- 1 APPARATUS AND METHOD OF PERFORMING MATRIX MULTIPLICATION OPERATION OF NEURAL NETWORK
- 1.1 Organization Name
- 1.2 Inventor(s)
- 1.3 APPARATUS AND METHOD OF PERFORMING MATRIX MULTIPLICATION OPERATION OF NEURAL NETWORK - A simplified explanation of the abstract
- 1.4 Simplified Explanation
- 1.5 Potential Applications
- 1.6 Problems Solved
- 1.7 Benefits
- 1.8 Potential Commercial Applications
- 1.9 Possible Prior Art
- 1.10 Original Abstract Submitted
APPARATUS AND METHOD OF PERFORMING MATRIX MULTIPLICATION OPERATION OF NEURAL NETWORK
Organization Name
Inventor(s)
Songyi Han of Hwaseong-si (KR)
APPARATUS AND METHOD OF PERFORMING MATRIX MULTIPLICATION OPERATION OF NEURAL NETWORK - A simplified explanation of the abstract
This abstract first appeared for US patent application 18397716 titled 'APPARATUS AND METHOD OF PERFORMING MATRIX MULTIPLICATION OPERATION OF NEURAL NETWORK
Simplified Explanation
The neural network apparatus described in the patent application is designed to perform matrix multiplication operations efficiently by determining whether to divide initial weights in a column or row direction based on reshape and transpose operations, generating division weights, performing matrix multiplication operations between input feature maps and division weights to generate intermediate feature maps, and ultimately producing a final feature map.
- The processor determines the direction in which to divide initial weights based on reshape and transpose operations.
- Division weights are generated by dividing the initial weight by a head count in the determined direction.
- Intermediate feature maps are generated by performing matrix multiplication operations between input feature maps and division weights.
- A final feature map is generated based on the intermediate feature maps.
Potential Applications
This technology can be applied in various fields such as image recognition, natural language processing, and signal processing.
Problems Solved
This technology solves the problem of efficiently performing matrix multiplication operations in neural networks by optimizing the division of initial weights based on reshape and transpose operations.
Benefits
The benefits of this technology include improved efficiency in matrix multiplication operations, which can lead to faster processing speeds and better performance in neural network applications.
Potential Commercial Applications
This technology can be utilized in industries such as healthcare (medical image analysis), finance (fraud detection), and autonomous vehicles (object recognition).
Possible Prior Art
One possible prior art for this technology could be the use of parallel processing techniques in neural networks to optimize matrix multiplication operations.
Unanswered Questions
How does this technology compare to existing methods for matrix multiplication in neural networks?
This technology optimizes the division of initial weights based on reshape and transpose operations, potentially leading to improved efficiency compared to traditional methods.
What impact could this technology have on the development of more advanced neural network models?
This technology could enable the development of more complex neural network models by providing a more efficient way to perform matrix multiplication operations, allowing for faster training and better performance.
Original Abstract Submitted
A neural network apparatus for performing a matrix multiplication operation includes a memory having at least one program stored therein and a processor to perform one or more operations by executing the at least one program. The processor can determine whether to divide an initial weight in one of a column direction and a row direction according to whether a reshape operation and a transpose operation are performed before or after a matrix multiplication operation and generate division weights by dividing the initial weight by a head count in the determined direction. Also, the processor can generate intermediate feature maps by performing a matrix multiplication operation between the input feature map and the division weights and generate a final feature map based on the intermediate feature maps.
