PyTorch自学笔记。学习教程为Zero to Mastery Learn PyTorch for Deep Learning，对应视频教程为https://www.youtube.com/watch?v=Z_ikDlimN6A
概念(What is deep learning)
机器学习(Machine learning, ML)
定义
将事物（数据）转化为数字，并找出数字中的模式
Machine learning is turning things (data) into numbers and find patterns in those numbers
（使用编程与数学方法使计算机完成对模式的寻找）
深度学习是机器学习的子集，而机器学习是人工智能的子集
概念理解
传统编程 vs 机器学习
以做菜为例：
传统学习：以原材料和制作步骤起始，制作出菜品
机器学习：以原材料和菜品起始，找出制作步骤
深度学习应用场景(What deep learning is good for or not good for)
应用场景
- 有大量规则的问题（无法手动列出所有规则）(Problems with long lists of rules)
- 不断变化的环境(Continually changing environments)
- 发现大量数据中的规律 (Discovering insights within large collections of data)
当可用简单的基于规则的系统解决问题时，不要使用机器学习
If you can build a simple rule-based system that doesn’t require machine learning, do that
非应用场景
- 需要解释的问题(When you need explainability)
- 传统方法更适用(When the traditional approach is a better option)
- 无法接受错误(When errors are unacceptable)
- 持有数据量不足(When you don’t have much data)
Machine learning vs Deep learning
应用场景
非深度学习的机器学习主要处理结构化数据(structured data)，如XGBoost
深度学习主要处理非结构化数据
常见模型
- Machine Learning
- Random forest
- Gradient boosted models
- Naive Bayes
- Nearest neighbour
- Support vector machine
- Deep Learning
- Neural Networks
- Fully connected neural network
- Convolutional neural network
- Recurrent neural network
- Tansformer