|
所在平台: Coursera |
课程主页: https://www.coursera.org/learn/advanced-deep-learning-with-pytorch
课程评论：没有评论
课程名称：使用PyTorch的深度学习 课程概述：本课程从基础的机器学习概念深入到使用PyTorch完成更复杂的深度学习模型和技术。内容涵盖了Softmax回归、浅层和深层神经网络以及卷积神经网络等特殊架构。 在本课程中，您将探索Softmax回归，并理解其在多类分类问题中的应用。您将学习如何训练神经网络模型，并探索相关知识。 课程大纲： 1. **逻辑回归与交叉熵损失**：了解均方误差的问题，讨论最大似然估计，并学习如何从最大似然估计计算交叉熵损失，最后使用PyTorch训练模型。您将在实验室中应用所学，并通过小测验测试您的概念。 2. **Softmax回归**：学习如何使用直线对数据进行分类，理解Softmax函数的工作原理，涵盖argmax函数及其用法。您将创建一个自定义的Softmax模块，并使用Softmax分类器创建分类模型，测试您的概念。 3. **浅层神经网络**：创建一个带有隐藏层的神经网络，学习如何训练神经网络模型及神经元如何提高模型性能。还将探讨过拟合、欠拟合、多类神经网络、反向传播和消失梯度，并在PyTorch中实现Sigmoid、Tanh和Relu激活函数。 4. **深层网络**：概述深度神经网络的实现，使用nn模块列表在PyTorch中实施深层神经网络，包括Dropout、层和权重的概念，探讨权重初始化问题及其解决方法，以及不同的初始化方法、梯度下降和批归一化。 5. **卷积神经网络**：描述卷积及激活图大小的确定，涵盖激活函数和最大池化，同时讨论多输入和输出通道的卷积。总结卷积神经网络构造器、前向步骤及在PyTorch中的训练，学习图形处理单元（GPU）、CUDA、残差网络及Resnet18。 6. **最终项目**：通过完成一个同伴评审的最终项目，展示并验证您在之前模块中所获得的技能。 本课程通过理论学习和实践实验结合，帮助学员深化对深度学习和PyTorch的理解与应用。
Name:Logistic Regression Cross Entropy Loss
Description:In this module, you will understand problem with mean squared error, and discuss maximum likelihood estimation. And then we'll see how to go from maximum likelihood estimation to calculating cross entropy loss, then Train the model PyTorch. You will apply your learnings in labs and test your concepts in quizzes.
Name:Softmax Regression
Description:In this module, you will learn how to use Lines to classify data and understand the working of the Softmax function. The module also covers the argmax function and its utilization. You will create a custom module for Softmax using the nn.module package in PyTorch and use a Softmax classifier to create a model for performing classifications. You will apply your learnings in labs and test your concepts in quizzes.
Name:Shallow Neural Networks
Description:In this module, you will create a neural network with a hidden layer using nn.Module and nn.Sequential. You will learn to train a neural network model and how neurons can improve a model. The model will also explain how to construct networks with multiple dimensional input in PyTorch. In addition, you will explore Overfitting and Underfitting, multi-class neural networks, back propagation and vanishing gradient. Finally, you will implement Sigmoid, Tanh and Relu activation functions in Pytorch. You will apply your learnings in labs and test your concepts in quizzes.
Name:Deep Networks
Description:This module provides an overview of deep neural network in Pytorch. You will learn to implement deep neural network in Pytorch using nn Module list. The module includes concepts like Dropout, layers, and weights. It will also discuss the problem of not initializing the Weights in a Neural Network model correctly and how to fix it. The module will also explore different initialization methods in Pytorch, gradient descent, and batch normalization. You will apply your learnings in labs and test your concepts in quizzes.
Name:Convolutional Neural Networks
Description:This module describes convolution and how to determine the size of the activation map. The module also covers activation functions and max pooling. In addition, the modaule discusses convolution with multiple input and output channels. It summarizes Convolutional Neural Network Constructor, Forward Step, and training in PyTorch. You will learn concepts like graphics processing units (GPUs), CUDA, residual network, and Resnet18. You will apply your learnings in labs and test your concepts in quizzes.
Name:Final Project
Description:In this module, you can complete a peer-reviewed final project to demonstrate and prove the skills you gained in the previous modules
This course advances from fundamental machine learning concepts to more complex models and techniques in deep learning using PyTorch. This comprehensive course covers techniques such as Softmax regression, shallow and deep neural networks, and specialized architectures, such as convolutional neural networks. In this course, you will explore Softmax regression and understand its application in multi-class classification problems. You will learn to train a neural network model and explore Ove
