Concept - Feature Extraction

Feature Extraction

Question: what properties of input $x$ might be relevant for predicting $y$? A feature extractor takes $x$ and outputs a set of (feature name, feature value) pairs (02 Pre-processing s.5).

Example for “ntu@gmail.com” (email-address task, s.4): Length>10 = 1; Length<50 = 1; contain ”@” = 1; endwith “com” = 1; endwith “sg” = 0; length between @ and . = 5; fraction of alpha = 0.85.

• Feature vector $\varphi (x)\in {\mathbb{R}}^d$: mathematically the names are dropped, only the values remain; the set of such vectors is the feature vector space (s.6, s.9).
• Weight vector $\mathbf{w}\in {\mathbb{R}}^d$: one parameter per feature = its contribution to the prediction (s.7); combined in a Concept - Linear Model.
• For text, features are like the columns of a spreadsheet; word occurrence is easy, grammatical function is hard (s.12).

Related

Concept - Linear Model | Concept - Text Preprocessing | Concept - Bag of Words

Appears in

• 02 Pre-processing - Chunk 01 (s.1-30)