Samsung electronics co., ltd. (20240303837). METHOD AND APPARATUS WITH CONVOLUTION NEURAL NETWORK PROCESSING simplified abstract
Contents
METHOD AND APPARATUS WITH CONVOLUTION NEURAL NETWORK PROCESSING
Organization Name
Inventor(s)
METHOD AND APPARATUS WITH CONVOLUTION NEURAL NETWORK PROCESSING - A simplified explanation of the abstract
This abstract first appeared for US patent application 20240303837 titled 'METHOD AND APPARATUS WITH CONVOLUTION NEURAL NETWORK PROCESSING
The abstract of the patent application describes a neural network apparatus that includes processors and a controller to determine shared operands for parallelized operations based on features of input feature maps and kernels.
- The neural network apparatus includes one or more processors and a controller.
- The controller determines a shared operand for parallelized operations, which can be a pixel value from input feature maps or a weight value from kernels.
- The shared operand is determined based on the features of the input feature maps and kernels.
- The processing units then perform the parallelized operations based on the determined shared operand.
Potential Applications: - Image recognition - Natural language processing - Autonomous driving systems
Problems Solved: - Enhancing the efficiency of neural networks - Improving the accuracy of machine learning models
Benefits: - Faster processing speeds - Enhanced performance in complex tasks - Reduced computational resources required
Commercial Applications: - Integration into hardware for AI applications - Development of specialized neural network processors
Questions about the technology: 1. How does the neural network apparatus improve the efficiency of parallelized operations? 2. What are the key features that differentiate this apparatus from existing neural network technologies?
Original Abstract Submitted
a neural network apparatus includes one or more processors comprising: a controller configured to determine a shared operand to be shared in parallelized operations as being either one of a pixel value among pixel values of an input feature map and a weight value among weight values of a kernel, based on either one or both of a feature of the input feature map and a feature of the kernel; and one or more processing units configured to perform the parallelized operations based on the determined shared operand.
