17964446. METHOD AND APPARATUS WITH NEURAL NETWORK OPERATION simplified abstract (Samsung Electronics Co., Ltd.)
Contents
METHOD AND APPARATUS WITH NEURAL NETWORK OPERATION
Organization Name
Inventor(s)
METHOD AND APPARATUS WITH NEURAL NETWORK OPERATION - A simplified explanation of the abstract
This abstract first appeared for US patent application 17964446 titled 'METHOD AND APPARATUS WITH NEURAL NETWORK OPERATION
Simplified Explanation
The abstract describes a method for implementing a neural network using a processor. Here is a simplified explanation of the abstract:
- The method involves generating a bit vector based on the input activations of the neural network.
- The bit vector is merged into the input activations and weights of the neural network, with the bit values becoming the most significant bits (MSBs) of multi-bit expressions.
- The input activations and weights are then sorted based on the MSBs.
- Finally, the neural network is implemented by performing operations between the sorted input activations and weights.
Potential applications of this technology:
- This method can be used in various fields where neural networks are applied, such as image recognition, natural language processing, and data analysis.
- It can improve the efficiency and speed of neural network implementations, making them more suitable for real-time applications.
Problems solved by this technology:
- Neural networks can be computationally intensive and require a lot of memory. This method addresses these issues by using bit vectors and sorting techniques to optimize the operations.
- By merging the bit vector into the input activations and weights, the method reduces the memory footprint and computational complexity of the neural network.
Benefits of this technology:
- The method improves the performance and efficiency of neural network implementations, allowing for faster and more accurate results.
- It reduces the memory requirements, making it possible to deploy neural networks on devices with limited resources.
- The optimized operations enable real-time processing, opening up possibilities for applications that require quick decision-making based on neural network analysis.
Original Abstract Submitted
A processor-implemented neural network method includes: generating a bit vector based on whether each of a plurality of input activations within a neural network is 0; merging the bit vector into the input activations such that bit values within the neural network included in the bit vector are most significant bits (MSBs) of multi bit expressions of the input activations; merging the bit vector into weights such that the bit values included in the bit vector are MSBs of multi bit expressions of the weights; sorting the input activations and the weights based on bits corresponding to the MSBs; and implementing the neural network, including performing operations between the sorted input activations and the sorted weights.
