Electronic device and control method thereof

CN114631100BActive Publication Date: 2026-07-07SAMSUNG ELECTRONICS CO LTD

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Patents(China)
Current Assignee / Owner: SAMSUNG ELECTRONICS CO LTD
Filing Date: 2020-11-03
Publication Date: 2026-07-07

AI Technical Summary

Technical Problem

Existing neural network models suffer from reduced computation speed when using complex activation functions to improve learning performance, while learning performance is reduced when using simple activation functions, making it difficult to balance computation speed and learning performance.

Method used

The activation function generation module in the electronic device generates new activation function information based on activation function information with high computational speed and high learning performance, and controls the switching of neural network models. It also combines gradient transformation functions to generate activation functions with high computational speed and excellent learning performance.

Benefits of technology

This technology enables simultaneous improvement in computational speed and learning performance in neural network models, enhancing the overall performance of the model without transmitting user privacy data.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN114631100B_ABST

Patent Text Reader

Abstract

An electronic device is provided. The device includes a communicator, a memory, and a processor. The memory is configured to store a neural network model, an activation function generation module that generates activation function information used in the neural network model, first activation function information, second activation function information, and third activation function information generated based on the first activation function information and the second activation function information. The processor is configured to receive fourth activation function information from an external device through the communicator, generate fifth activation function information by inputting the first activation function information and the fourth activation function information to the activation function generation module based on type information about the fourth activation function information corresponding to the second activation function information, and control the neural network model to change the activation function information used in the neural network model from the third activation function information to the fifth activation function information.

Need to check novelty before this filing date? Find Prior Art