Deep Learning Notes: Chapter 1 Introduction
前言
最近開始讀《Deep Learning》一書。這讓我有了一個邊讀書邊寫筆記的動機：能夠讓人很輕鬆流暢的把握住這本書的脈絡，從而讀懂這本書的核心內容。
由於終究是英文表達更地道，因此該筆記都是節選自書中的原文。只有在我比較有把握的情況下才會給個別概念加上中文翻譯。另外，“個人總結”部分是我自己的總結。各位讀者如果有建議或意見，歡迎留言。謝謝！
Deep Learning Chapter 1 Introduction
| Concept | Chinese | Description |
|---|---|---|
| Artificial Intelligence (AI) | 人工智慧 | Intelligent software to automate routine labor, understand speech or images, make diagnoses in medicine and support basic scientific research. |
| Machine Learning | 機器學習 | AI systems acquire their own knowledge by extracting patterns from raw data. |
| Representation Learning | 表示學習 | Use machine learning to discover not only the mapping from representation to output but also the representation itself. |
| AI Deep Learning | AI 深度學習 | Computers learn from experience and understand the world in terms of a hierarchy of concepts. |
In the early days of AI, the field rapidly tackled and solved problems that are intellectually difficult for human beings but relatively straightforward for computers — problems that can be described by a list of formal (形式化), mathematical rules.
Reason: Abstract and formal tasks that are among the most difficult mental undertakings for a human being are among the easiest for a computer.
Challenge to AI: Problems that human solve intuitively, but hard to describe formally.
Example: Recognizing spoken words or faces in images.
Key challenge: How to get informal (非形式化) knowledge into a computer.
A solution: Machine learning.
Challenge to simple machine learning: The performance of simple machine learning algorithms depends heavily on the representation of the data they are given.
Key challenge: What features should be extracted. Feature (特徵): The piece of information included in the representation.
A solution: Representation learning.
Goal of representation learning: To separate the factors of variation that explain the observed data. Factors: Sources of influence, can be thought of as concepts or abstractions that help us make sense of the rich variability of the data.
Challenge to representation learning: Disentangle the factors of variation and discard the ones that we do not care about.
A solution: Deep learning.
Method: Introducing representations that are expressed in terms of other, simpler representations.
Two main ways of measuring the depth of a mode:
1. The depth of the computational graph.
2. the depth of the graph describing how concepts are related to each other. It is used by deep probabilistic models,
個人總結
| 概念 | 輸入 | 輸出 |
|---|---|---|
| Simple machine Learning | 特徵 | 最終結果 |
| Representation Learning | 原始資料 | 特徵 |
| Deep Learning | 原始資料 | 多層次特徵，就像一棵樹，上一層特徵是下一層特徵的抽象。下一層特徵更簡單。 |