Week 1 - Elements of Machine Learning Algorithms
Definition
The Mathematically Definition of Machine Learning:
$A:S \in (X \times Y)^n \mapsto h_s \in H$
Input training data: $S = {(X_1,Y_1),…,(X_n,Y_n)}$
Input predefined hypothesis class: $H={h_1,h_2}$
The objective function and optimization method together make up a mapping: $A:(x \times y)^n \mapsto H$
Output hypothesis: $h_s$
The overall learning algorithm is mapping: $A:S \in (X \times Y)^n \mapsto h_s \in H$
Input data
Noise in data:
- Feature Noise
- Label Noise
Input Predefined Hypothesis Class
How to choose the hypothesis class?
- Data analysis
- Complexity
- Cross-Validation
- Prior knowledge