Models and Algorithms

From Machine Learning
Jump to: navigation, search

Algorithms

An algorithm is a mathematical technique or equation where the parameters have not been assigned values. It is a non-specific framework.

Example: y = mx + b

At this stage, the algorithm does not have parameters, but it may have hyperparameters. A solution proposal could include an algorithm alongside specified hyperparameters such as the number of k-clusters or the learning rate of a neural network.

Models

A model is an algorithm with fully defined parameters as learned from the data. Iteration will refine the values as the code is run.

Example: y = 0.11x + 0.5

Algorithms become models once parameter values are assigned. Again, hyperparameters exist separately to this definition.


In practical terms, a model is how the layers of data are organised.

Algorithm vs model.png

Coding Deep Learning for beginners is a good resource for starting out learning these definitions.