What is an algorithm?
An algorithm is a precise, finite sequence of steps for turning an input into an output. A recipe for sorting a list of invoices by due date is an algorithm, and so is the procedure a routing tool uses to order delivery stops. In machine learning the word appears twice. A training algorithm, such as gradient descent, adjusts a model’s parameters to fit the data. The trained model is then applied by an inference algorithm that computes the prediction for each new input.
Public discussion often uses “the algorithm” to mean a whole AI system, as in a lending algorithm or a social media feed. Technically the algorithm is the recipe and the model is what the recipe produced from data. The distinction matters when something goes wrong: a biased lending decision usually traces to the training data or the model rather than to the arithmetic of the algorithm itself.
Why an algorithm matters
Algorithms differ in speed, memory use, and whether they are deterministic (the same input always produces the same output) or probabilistic. A warehouse optimization problem with millions of possible pick paths needs an algorithm that finds a good answer in seconds rather than the perfect answer in days. When a vendor describes a proprietary algorithm, ask what problem it solves, what it is guaranteed to produce, and how its results were validated on data like yours.