Features

Features are quantitative, measurable numerical attributes or learned vector embeddings extracted from raw input entities that serve as explanatory variables for model computation.

Real-World Mental Model

“The specific clues a doctor measures (blood pressure, temperature, heart rate) to determine a patient’s diagnosis.”

INTERACTIVE ENGINE: FEATURES VECTOR

Live Reactive

Select Real-World Entity to Project:

Dense Vector Representation

vec(Cat) = [0.85, 0.25] ∈ &Ropf;²

Biological / Small

Dimension 1: Animacy

+0.85 (Biological)

Dimension 2: Scale/Weight

0.25 (Normalized Mass)

CORE PRINCIPLES

What You Must Understand

Tabular features: Age, income, transaction amount, latitude.

Text features: Sub-word token IDs, dense 1536-dimensional embeddings.

Feature normalization (scaling to [0, 1] or Z-score standard deviation) prevents large values from dominating gradients.

Reference Implementation

// Feature Normalization (Min-Max)
const normalized = (x - min) / (max - min);

Production Failure Mode

Feeding unnormalized features with radically different scales (e.g. income in $100,000s vs age in 10s) into gradient-based optimizers.

View Full 26-Topic Curriculum Table

Loading AI Engineering Foundations...