18464044. COMPILATION METHOD AND APPARATUS WITH NEURAL NETWORK simplified abstract (SAMSUNG ELECTRONICS CO., LTD.)
COMPILATION METHOD AND APPARATUS WITH NEURAL NETWORK
Organization Name
Inventor(s)
COMPILATION METHOD AND APPARATUS WITH NEURAL NETWORK - A simplified explanation of the abstract
This abstract first appeared for US patent application 18464044 titled 'COMPILATION METHOD AND APPARATUS WITH NEURAL NETWORK
Simplified Explanation
The abstract describes a method for compiling a neural network, which involves grouping layers, generating passes and threads, and performing compilation in parallel.
- The method receives data related to the neural network.
- The layers in the neural network are grouped based on the data.
- A set of passes is generated, which can be executed in parallel to process the neural network.
- A set of threads is generated to perform optimization functions independently or sequentially based on layer dependencies.
- Compilation is performed in parallel based on the grouped layer, passes, and threads.
Potential Applications
- This method can be applied in various fields where neural networks are used, such as image recognition, natural language processing, and autonomous vehicles.
- It can be used in research and development of new neural network architectures and models.
- The method can be integrated into existing neural network frameworks and libraries to improve their performance and efficiency.
Problems Solved
- The method addresses the challenge of optimizing the compilation process for neural networks, which can be computationally intensive and time-consuming.
- It solves the problem of efficiently processing and optimizing neural networks with multiple layers and complex dependencies.
- The method helps to improve the overall performance and efficiency of neural network computations.
Benefits
- By grouping layers and executing passes in parallel, the method reduces the overall compilation time for neural networks.
- The ability to perform optimization functions independently or sequentially based on layer dependencies allows for more efficient utilization of computational resources.
- The method improves the performance and efficiency of neural network computations, leading to faster and more accurate results.
Original Abstract Submitted
A compile method for a neural network, the compile method includes receiving data related to the neural network, generating a grouped layer by grouping layers comprised in the neural network based on the data, generating a set of passes executable in parallel based on a dependency between a plurality of passes to process the neural network, generating a set of threads performing a plurality of optimization functions based on whether optimization operations performed by the optimization functions is performed independently for the layers, respectively, or sequentially based on a dependency between the layers, and performing compilation in parallel based on the grouped layer, the set of passes, and the set of threads.
