17887183. METHOD AND APPARATUS FOR NEURAL NETWORK OPERATION simplified abstract (Samsung Electronics Co., Ltd.)

METHOD AND APPARATUS FOR NEURAL NETWORK OPERATION

Organization Name

Samsung Electronics Co., Ltd.

Inventor(s)

Donghyun Lee of Seongnam-si (KR)

Joonsang Yu of Seoul (KR)

Junki Park of Suwon-si (KR)

Jungwook Choi of Seoul (KR)

METHOD AND APPARATUS FOR NEURAL NETWORK OPERATION - A simplified explanation of the abstract

This abstract first appeared for US patent application 17887183 titled 'METHOD AND APPARATUS FOR NEURAL NETWORK OPERATION

Simplified Explanation

The abstract describes a neural network operation apparatus that performs a non-linear function in a neural network operation. Here is a simplified explanation of the abstract:

• The apparatus receives input data for the neural network operation.
• It uses a quantized Look Up Table (LUT) that corresponds to a non-linear function in the neural network operation.
• The input data is scaled up based on a scale factor.
• A quantized LUT parameter is extracted from the quantized LUT using the scaled-up input data.
• The apparatus generates an operation result by performing the neural network operation based on the quantized LUT parameter.

Potential applications of this technology:

• Artificial intelligence and machine learning systems
• Image and speech recognition systems
• Natural language processing systems
• Autonomous vehicles and robotics

Problems solved by this technology:

• Efficiently performing non-linear functions in neural network operations
• Reducing the computational complexity of neural network operations
• Improving the speed and efficiency of neural network operations

Benefits of this technology:

• Faster and more efficient neural network operations
• Reduced computational resources required
• Improved accuracy and performance of artificial intelligence systems

Original Abstract Submitted

A neural network operation apparatus may include a receiver configured to receive input data to perform the neural network operation and a quantized Look Up Table (LUT) corresponding to a non-linear function comprised in the neural network operation, and a processor configured to perform scale-up on the input data based on a scale factor, to extract a quantized LUT parameter from the quantized LUT based on scaled-up input data, and to generate an operation result by performing a neural network operation based on the quantized LUT parameter.