NTU Machine Learning - Lec3

這講建構了一個大略的 ML 世界觀。

Different Output Space

1. Binary classification ($|\mathcal{Y}| = 2$)
2. Multiclass classification ($|\mathcal{Y}| = K$)
3. Regression ($\mathcal{Y} \subset \mathbb{R}$)

Different Label

1. Supervised (w/ explicit label)
2. Unsupervised (e.g clustering)
3. Semi-supervised (combine above 2, owing to label cost issue)
4. Reinforcement (learn partial info. sequentially)

Different Protocol

1. Batch learning (Given all known data once)
2. Online learning (can improve when seeing new data)
3. Active learning (interactively query label $y_n$ of chosen $x_n$)，也是很常見的蒐集 data 方式(欸！？)，跟 Online Learning 類似，但差別在會 主動 query

Different Input Space

1. Concrete feature (利用human intelligence 抽取出有關連的特徵將 input 做簡化)
2. Raw feature (通常需要再做 feature engineering，但現今 DL 的前半部通常會一起 end-to-end train 成一個還不錯的 feature extractor)
3. Abstract feature (原先的 feature 沒有任何物理意義(e.g 可能只是 ID 而已)，需要做 feature conversion 轉成上述二者)

Reference

• 林軒田老師的講義

comments powered by Disqus