相关实验视频
Updated: Jun 9, 2025

Deep Neural Networks for Image-Based Dietary Assessment
Published on: March 13, 2021
分数级随机梯度下降方法与动量和能量用于深度神经网络
Xingwen Zhou1, Zhenghao You2, Weiguo Sun2
1School of Information Science and Engineering, Lanzhou University, 222 Tianshui South Road, Chengguan District, Lanzhou, Gansu Province, Lanzhou, 730000, Gansu, China; School of Nuclear Science and Technology, Lanzhou University, 222 Tianshui South Road, Chengguan District, Lanzhou, Gansu Province, Lanzhou, 730000, Gansu, China.
一种新的分数级随机梯度下降与动量和能量 (FOSGDME) 方法改善了图像分类. 这种新的方法提高了趋同性和准确性,在CIFAR-10数据集上表现优于传统的优化算法.
科学领域:
- 机器学习 机器学习
- 优化算法 优化算法
- 分数微积分的计算.
背景情况:
- 现有的分数梯度算法面临的挑战是汇聚到真正的极端点.
- 需要在深度学习中改进优化方法,以提高准确性和速度.
研究的目的:
- 提出一种新的分数级随机梯度下降与动量和能量 (FOSGDME) 方法.
- 解决分数梯度方法中的收问题.
- 提高优化算法的收速度,准确性和稳定性.
主要方法:
- 修改的卡普托分数顺序导数定义,用于分数顺序的随机梯度下降 (FOSGD).
- 集成的势头信息来开发具有势头的FOSGD (FOSGDM).
- 引入能量形成以产生具有动量和能量的FOSGD (FOSGDME).
主要成果:
- 与整数顺序优化相比,FOSGD,FOSGDM和FOSGDME算法表现出更高的性能.
- 在CIFAR-10图像分类数据集上取得了最先进的结果.
- 使用ResNet和DenseNet架构进行实验验证.
结论:
- 拟议的FOSGDME方法为深度学习任务的优化提供了显著的改进.
- 新的分数顺序方法为传统优化提供了强大的,准确的替代方案.
- 超出性能验证了将动量和能量纳入分数优化中的有效性.
更多相关视频
相关概念视频
Gradient and Del Operator
Moment-of-Momentum Equation
Linear Momentum in Control Volume
One-Degree-of-Freedom System
A one-degree-of-freedom system is defined by an independent variable that determines its state and behavior. One example of a one-degree-of-freedom system is a simple harmonic oscillator, such as a...
Impulse-Momentum Theorem
By writing Newton's second law of motion in terms of the momentum of an object and the external force acting on it, and simultaneously using the definition of the impulse vector, it can be shown that the total impulse on an object is equal to its...
Conservation of Linear Momentum for a System of Particles
The impulsive force at play during this interaction is of extremely short duration, rendering its impulse negligible. When...

