Supervised vs. Unsupervised Learning

何謂機器學習 (Machine Learning)?

• 透過演算法將已知推論到未知。 "Algorithms for inferring unknowns from knows"

• 應用場景包含： 圖像、聲音識別; 預測; 導航

探討機器學習前，一般會將處理的問題分成 Supervised vs. Unsupervised Learning 兩類：

• Supervised: Given the data $(x^i,y^i)$, choose a function $f(x^i)=y^i$.

  • Classification: $y_i \in $ { $ Finite~Set $ }$ $
  • Regression: $y_i \in \mathbb{R}$
    Note: $x^i \in \mathbb{R}^2$ which means $x^i$ equals to $(x^i_1,x^i_2)$

• Unsupervised: Given the data $(x^i,..., x^n)$, find the pattern in the data.

  • Clustering
  • Density Estimation
  • Dimensionality Reduction

這只是比較概略的劃分，事實上還有其他更細緻的方法，如：

• Semi-supervised Learning: Given the $(x^1,y^1),...,(x^k,y^k),x^{k+1},...,x^n$, predict $y^{k+1},...,y^n$.
• Active Learning
• Decision Theory
• Reinforcement Learning

自己因為碩論的關係，關注議題以 Reinforcement Learning Model 為主，但在實際運用上比較常見的機器學習屬於Supervised, Unsupervised 和 Semi-supervised Learning.