📝 Model Families & Architectures

Description

Purpose: model families and architectures describe the kinds of models and structures used to represent relationships and compute outputs. Examples include linear models, probabilistic models such as Naive Bayes, decision-tree ensembles, and neural networks such as convolutional networks and Transformers.
Scope: model families group related approaches, while architectures specify how computational components are arranged. Foundation models are grouped here by broad pretraining and adaptability across tasks; they can also belong to a neural-network family. These groupings overlap rather than forming mutually exclusive classes.

Linear Regression models a numerical target using a linear relationship.
Naive Bayes uses probabilistic assumptions for classification.
Decision Trees make predictions through feature-based splits.
Neural Networks compose trainable computational layers.
Next-Concept Modeling (NCM) generates text through a learned hierarchy of discrete codes.
Learning Paradigms describes how models are trained.