AI glossary: terms starting with A

A parameter with one star, which collects extra positional arguments into a tuple: def total(*prices).

size of observed duration minus forecast, sign ignored.

The size of the residual.

A name that collects a result across a loop's iterations. It is created before the loop (total = 0), updated in the body (total += steps), and read after it.

The fraction of answers that are right: (spam caught + ham passed) / all.

an inequality that holds with equality at the optimum

Example: x + y ≤ 1 at (1, 0)

A declared interval of recorded ages.

The span of source dates combined into one reported value.

A second name bound to the same object: after backup = cart, backup is an alias of cart.

Alanine aminotransferase, a laboratory measurement reported in U/L.

cos θ = a · b / (‖a‖ ‖b‖)

Example: 26.57°

A person who labels examples.

Finding unusual examples (Module 13).

a function whose derivative is the given function

Example: x³/3 for x²

Approved count divided by the stated group considered by the rule.

Position of the maximum, not the maximum value

the input that gives the lowest, or highest, output

Example: argmin (w − 3)² + 1 = 3

A value passed to a function. In print("Total:", 20.75), the two arguments are "Total:" and 20.75.

Values in a typed buffer, interpreted through a shape

A statement that a condition is True; assert raises AssertionError when it is not.

The instruction to give a value a name: total = 24.75. The = does not mean "equal"; it tells Python to evaluate the right hand side and bind the name on the left to it.

name := value: assigns a value and gives that value, inside an expression. while (price := input("Price: ")) != "": reads and tests in one condition. Also called the walrus operator.

Recorded values vary together in the observed data.

Uncertainty about whether model inputs and rules describe the real decision; more runs do not reduce it.

A value stored on an object, reached with a dot: review.rating. vars(review) shows an object's attributes as a dictionary.

the area under the ROC curve

Example: 0.7656 on the tiny ward

Association between a series and a lagged copy of itself.

The expected operation count under a stated distribution of inputs. For n distinct values with equally likely successful targets, linear search averages (n + 1) / 2 comparisons. Duplicates or unequal target probabilities change this mean.

Dimension named by an operation; reduction removes that dimension

The value at the start of a chart axis.

one interval per feature; the region a tree's splits describe

Example: 12 of 14 on the ward