PyTorch-26H-1
PyTorch-26H-1
主页：https://www.freecodecamp.org/news/learn-pytorch-for-deep-learning-in-day/
youtub：https://youtu.be/V_xro1bcAuA
github：https://github.com/mrdbourke/pytorch-deep-learning
Learn PyTorch for Deep Learning: Zero to Mastery book：https://www.learnpytorch.io/
PyTorch documentation：https://pytorch.org/docs/stable/index.html
Chapter 0 – PyTorch Fundamentals
what is deep learning?
Machine learning is turning things (data) into numbers and finding patterns in those numbers.
Deep Learning ∈ Machine Learning ∈ Aritfical Intelligence
传统程序：输入+规则→输出
机器学些：输入+输出→规则
Why use machine/deep learning?
对于复杂的问题，无法找到所有的规则。
The number one rule of ML
“If you can build a simple rule-based system that doesn’t require machine learning, do that.”
A wise software engineer… (actually rule 1 of Google’s Machine Learning Handbook)
What deep learning is good for
Problems with long lists of rules- when the traditional approach fails, machine learning/deep learning may help.
规则列表过长的问题——当传统方法失败时，机器学习/深度学习可能会有所帮助。
Continually changing environments- deep learning can adapt (learn’) to new scenarios.
不断变化的环境——深度学习可以适应（学习）新场景。
Discovering insights within large collections of data- can you imagine trying to hand-craft rules for what 101 different kinds of food look like?
在大量数据中发现见解——你能想象尝试手工制定 101 种不同食物的规则吗？
- What deep learning is not good for
When you need explainability- -the patterns learned by a deep learning model are typically uninterpretable by a human.
当你需要可解释性时——深度学习模型学习到的模式通常无法被人类解释。
When the traditional approach is a better option一if you can accomplish what you need with a simple rule-based system.
当传统方法是更好的选择时——如果你可以使用简单的基于规则的系统完成所需的工作。
When errors are unacceptable一since the outputs of deep learning model aren’t always predictable.
当错误不可接受时——因为深度学习模型的输出并不总是可预测的。
When you don’t have much data一deep learning models usually require a fairly large amount of data to produce great results.
当你没有太多数据时——深度学习模型通常需要相当大量的数据才能产生很好的结果。
Machine learning vs deep learning
- Machine learning
适合处理结构化数据
常见算法：
Random forest 随机森林
Gradient boosted models 梯度提升模型
Naive Bayes 朴素贝叶斯
Nearest neighbour 最近邻
Support vector machine 支持向量机
- Deep learning
适合处理非结构化数据
常见算法：
Neural networks 神经网络
Fully connected neural network 全连接神经网络
Convolutional neural network 卷积神经网络
Recurrent neuralnetwork 循环神经网络
Transformer
Anatomy of neural networks
数据→数字→神经网络→权重→输出
Different learning paradigms
监督学习：大量已知数据标注。
无监督学习：自动分析数据。
迁移学习：将学习到的模式嵌入到新的模型中。
强化学习reinforcement learning：奖励想要的结果。
What can deep learning be used for?
