18604268. NEURAL NETWORK ACCELERATOR simplified abstract (SAMSUNG ELECTRONICS CO., LTD.)
Contents
NEURAL NETWORK ACCELERATOR
Organization Name
Inventor(s)
Hyungjun Kim of Pohang-si (KR)
Jae-Joon Kim of Pohang-si (KR)
NEURAL NETWORK ACCELERATOR - A simplified explanation of the abstract
This abstract first appeared for US patent application 18604268 titled 'NEURAL NETWORK ACCELERATOR
The neural network accelerator described in the patent application includes various components such as bit operators, adders, shifters, and accumulators to efficiently process input feature data and weight data.
- First bit operator performs multiplication on specific bits of input feature data and weight data to generate a first multiplication result.
- Second bit operator performs multiplication on different bits of input feature data and weight data to generate a second multiplication result.
- Adder combines the first and second multiplication results to generate an addition result.
- Shifter shifts the addition result by a specified number of digits based on a shift value to produce a shifted addition result.
- Accumulator generates output feature data based on the shifted addition result.
Potential Applications: - This technology can be used in various AI applications such as image recognition, natural language processing, and autonomous driving systems. - It can also be applied in hardware accelerators for deep learning tasks in data centers and edge devices.
Problems Solved: - Improves the efficiency and speed of neural network computations. - Reduces the computational complexity of multiplying and adding large sets of data.
Benefits: - Faster processing of neural network operations. - Lower power consumption in AI hardware accelerators. - Enhanced performance in AI applications.
Commercial Applications: Neural network accelerators using this technology can be integrated into AI processors for smartphones, smart cameras, and IoT devices, enhancing their AI capabilities and performance.
Questions about Neural Network Accelerators: 1. How does this technology improve the efficiency of neural network computations? 2. What are the potential commercial applications of neural network accelerators using this innovation?
Original Abstract Submitted
Disclosed is a neural network accelerator including a first bit operator generating a first multiplication result by performing multiplication on first feature 5 bits of input feature data and first weight bits of weight data, a second bit operator generating a second multiplication result by performing multiplication on second feature bits of the input feature data and second weight bits of the weight data, an adder generating an addition result by performing addition based on the first multiplication result and the second multiplication result, a shifter shifting a number 10 of digits of the addition result depending on a shift value to generate a shifted addition result and an accumulator generating output feature data based on the shifted addition result.
