|
所在平台: Udemy |
课程主页: https://www.udemy.com/course/cnns-with-mathematical-derivations-practical-application/
课程评论：没有评论
课程名称：大学级神经网络 [II] - 卷积网络：数学与实践！ 课程概述：本课程是《神经网络与深度学习》系列的第二部分，前一部分为“大学级神经网络与数学推导”。本课程专注于卷积神经网络（CNN），这是一种主要用于图像和视频视觉识别的特殊神经网络。我们将重点讨论卷积神经网络的概念、直觉、数学推导及其实际应用。 课程分为四个章节： 第一章着重于CNN的基本概念和直觉，包括什么是CNN？它们如何工作？为何适合视觉识别等基础问题。 第二章深入探讨CNN的数学推导，介绍前向和反向传播方程的推导过程，以及如何根据超参数（如卷积核大小、步幅和池化）变化而变化。 第三章关注不同类型的卷积和池化技术，探讨适用于各种任务的概念，包括3D卷积、扩张卷积、全局池化、点卷积、空间和深度可分卷积、反卷积、分组卷积、随机卷积等，同时提供何时适用和不适用这些技术的指导和见解。 第四章专注于依赖于CNN的各种实际应用。与其简单总结几个应用中的关键概念和算法，我选择对几篇高质量研究论文进行深入阅读，逐段分析，并解释任何不清晰的概念或方程。我们将对比不同论文的研究方法和结果，使这一章节保持动态和扩展。 第四章的主要目的并非教授特定算法，而是培养读懂研究论文的能力，理解不同论文之间的关系与引用，以及研究者如何通过技巧提升性能。这种思维方式对任何希望进入深度学习领域的研究者或工程师都将有巨大益处。 希望你喜欢这门课程，觉得它有用！期待在下一个视频再见！
Convolutional neural networks with mathematical derivations and practical applications is the second course in my Neural Networks and deep learning series, after the first course in the series named "College-Level Neural Networks With Mathematical Derivations".As the title implies, This course is focused on Convolutional neural networks, a special kind of neural networks mainly used for visual recognition in images and videos, yet not limited to that.In this course, I mainly focus on concepts, intuitions, mathematical derivations, and practical applications. The course is mainly divided into 4 chapters: Chapter 1 focuses on the conceptual basics and intuitions of CNNs. What are CNNs ? How do they operate? Why are they suitable for visual recognition ? and so on?Chapter 2 takes a step deeper into the CNN mathematical derivations. What are forward and backward propagation equations through CNNs? How are they derived ? How do they change with changes in hyperparameters like kernel sizes, strides, and pooling? Chapter 3 takes a step higher and focuses on different types of convolutions and pooling suitable for various tasks. Ideas like 3D convolutions, dilated convolutions, global pooling, pointwise convolutions, spatial and depth-wise separable convolutions, deconvolutions, grouped convolutions, shuffled convolutions and more are covered in detail, along with justifications and insights on when to and not to use them in practice.Moving on to Chapter 4, I decide to take an even larger step higher and focus on practical applications that depend heavily on CNNs. My way of handling this is different. Instead of just summarizing a few key ideas and algorithms used for a couple of different applications on a very high level, I opt for diving very deeply and extensively in a couple of chosen high-quality research papers that introduce a specific algorithm or idea.For each paper, we read its paragraphs together, line by line, and I explain any unclear concepts or equations as we proceed. We move from one paper to another, comparing their approaches and results. This chapter is designed to be an ever-growing, dynamic chapter.The main purpose of Chapter 4 is NOT to teach the specific algorithms presented. In fact, new algorithms emerge every few months anyway rendering older algorithms nearly obsolete.Rather, the goal is to get a feel of research papers, how to read them and understand them, and how different research papers relate to each other, reference each other and build upon each other's work. How researchers introduce a lot of tips and tricks to raise their performance and how they justify such choices.Growing up this mindset will have a huge benefit for anyone who wishes to enter the deep learning field, either as a researcher or an engineer.Hope you enjoy the course and find it useful! See you, in the next video!
