Model fitting for classification (1): problem statement (deterministic version)

Antonio Sala, UPV

Difficulty: * ,       Relevance: ,      Duration: 16:28

*Enlace a Spanish version

Summary:

This video poses the problem of fitting some parameters $𝜃$ in a model $f(x,𝜃)$ to fit pairs of training data samples $(x_i,y_i)$ with $y_i\in \{0,1\}$, that is, binary classification (supervised). The problem can be deterministic (dog/not-dog) or random (probability of having a certain genetic mutation when some marker appears in a blood test). This video discusses the deterministic problem, presenting the issues that arise when “perfect” classification is possible, and the cost functions for “imperfect” classification problems where false positives or false negatives are inevitable.

The probabilistic version of the statement of this type of problems is addressed in the video [clasifintr2EN].

*Link to my [ whole collection] of videos in English. Link to larger [ Colección completa] in Spanish.