Neural networks are a way to model data using a combination of affine and non-linear transformations.

Regression & Classification

  • Regression — Predicting values from given input values. Returns continuous values. If more than one value is returned, it is called multivariate regression.
  • Classification — Predicting a category. Returns discrete categories. If prediction is from one of two categories, it is called binary classification. If prediction is from more than two categories, it is called multi-class classification.

Network Structure

Research suggests that generalization is associated with well-structured representations. Different types of network structures exist depending on the type of problem. For example CNNs use a specific neural network structure to process images.


This page is a part of Machine Learning.