AI glossary: terms starting with S

a point with zero gradient that is neither a minimum nor a maximum

Example: the origin of x² − y²

observed recorded trips.

Average centred cross-product, using one less than the sample count.

Sum of sampled numeric values divided by sample size.

The outcomes under consideration; here, all 284,807 recorded transactions.

The square root of the sample variance, in the original unit.

A number calculated from the selected sample, such as its mean.

The sum of squared deviations divided by n - 1. Its unit is the original unit squared.

Distribution of a statistic across repeated samples under one design.

The operational set or process from which sample members can be selected.

One eligible invoice ID, the unit a draw selects.

One ID appears at most once in a draw; the full frame is available again for the next draw.

every component times one number

Example: 2 · (3, 1) = (6, 2)

One point per paired numerical observation.

Field names and stored data types in a table, as df.info() shows them.

a number times a power of ten; Python writes it 3.6e-06

Example: 3.6 × 10⁻⁶

The region of a program in which a name is available; a comprehension's loop variable has its own scope.

w · x + b for a point x

Example: 6 for P = (4, 3)

The number of fits a search may use.

a line through two points of a curve

Example: through (1, 1) and (2, 4), slope 3

the derivative of the derivative, f″(x)

Example: x³ − 3x → 6x

Starting value used to reproduce a generator's sequence

A rule determining which source rows enter a reported subset.

When the way rows are chosen makes the data differ from the population a claim is about.

A different sample choosing different features.

Showing only the results that found something, and leaving out the checks that did not.

The conventional first parameter of an instance method: the object the method was called on. review.is_positive() is Review.is_positive(review).

Learning from labels taken from the data itself, as a language model predicts the next word.

Learning from a few labelled and many unlabelled examples.

Recompute after one stated change to see its effect.

A value that marks the end of the input, and is not processed, such as an empty entry at a checkout.

A special value that stands for a missing or unusual case, not an amount: pdays = 999.

Text used between fields when splitting or joining strings.

An unordered collection in which each value appears at most once. An empty set is set().

A set built from an expression evaluated for each item: {len(item) for item in items}.

A new outer collection that still refers to the original elements; nested mutable elements remain shared.

Element count on each axis, as a tuple

A number defining an assumed Beta curve before sampled evidence.

the matrix's number of columns must equal the vector's length; the result has one entry per row

Example: (3, 2) @ (2,) → (3,)

(m, k) @ (k, n) → (m, n)

Example: (2, 3) @ (3, 4) → (2, 4)

adding k, or replacing x with x − h

Example: (x − 3)² + 1

and and or stop as soon as they know the answer: in False and x and True or x, the x is never evaluated.

Pulling the weights toward zero.

An action such as printing or mutating an object, beyond producing a return value.

1 / (1 + e⁻ˣ), an S-shaped curve between zero and one

Example: 0.5 at x = 0

negative when the transformation flips the plane over

Example: the flip, −1

Information a feature carries about the label; the less the class distributions overlap, the more signal.

A function's name and its parameters, as the documentation lists them: sorted(iterable, /, *, key=None, reverse=False).

(w · x + b) / ‖w‖, positive on the side w points to

Example: 2.6833

the binary digits of the number, 53 of them in float64

Example: 1.1₂ for 0.75

a rule, calculated from recorded trips, that gives a prediction for a new trip. Both forecasts in 00.1, one duration for every trip and one duration per pickup zone, are simple models.

Every set of this size has equal chance.

An overall comparison that reverses inside the subgroups because the groups have different compositions.

An explicit teaching rule or process applied to data and labelled as such; here the screen PAY_0 <= 0, not an actual lender rule.

a square matrix with no inverse; it flattens at least one direction, and its determinant is zero

Example: [[1, 2], [2, 4]]

the length of a semi-axis of the ellipse; how much the matrix stretches one direction

Example: 6.7082, 2.2361

vᵢ, the input directions, and uᵢ, where they land

Example: v₁ at 45°, u₁ at 71.57°

A distribution with a long tail on one side, whose few large values pull its mean away from its median: area_se, median 24.61, largest 542.2.

A range of elements, from a start up to but not including a stop, with an optional step: temps[2:5], temps[::-1]. Slicing a list creates a new outer list; slicing a tuple or string returns the same type, possibly reusing an unchanged immutable object.

The spacing and direction used to select positions in a slice.

Change in fitted minutes per additional mile.

Replacing nearby values with a declared local summary, such as a three-week moving average.

Shrinking a value toward zero by α and setting it to zero when it is within α of zero.

The recorded cases a summary begins with: 30,000 existing clients.

Systematic difference between the available frame and the population a claim is meant to describe.

How memory use grows with input size. Extra storage excludes the input; auxiliary space also excludes the returned output. State which is being counted.

every combination of a set of vectors

Example: a plane through the origin

Many recorded zeros with fewer nonzero entries.

Many weights exactly zero.

A method with two underscores on each side of its name, which Python calls itself: __init__, __repr__, __eq__, __add__.

A curve built from joined low-degree pieces.

How far observed values extend around a stated center.

The square of an observed value minus the group mean.

A residual multiplied by itself.

A sort that keeps items with equal keys in their original order. sorted and list.sort are stable.

Outcome segments placed end to end within a group. Scaled to equal height, it shows fractions and hides group size.

Square root of variance, in the original units: about $306.02 for the first rule's costs.

Spread of a statistic across repeated samples.

The fold scores' standard deviation over the square root of the number of folds.

The modules that come with Python: math, pathlib, csv, json, importlib.

Normal model with mean zero and SD one.

The change in the prediction for a one-standard-deviation change in the feature.

subtract the mean, divide by the standard deviation

Example: the lot's age and mileage

Recomputing group rates under one declared subgroup mix.

the optimal value for an objective

Example: w* = (2.7667, −0.3333)

One instruction. Python executes a program's statements one at a time, from top to bottom.

a point where the derivative is zero

Example: x² − 3x + 2 at 1.5

reasoning from observed cases about unobserved cases, such as using March trips to choose a forecast for trips that have not happened yet.

describing observed data and reasoning about unobserved cases.

Each update uses one training row, in shuffled order.

The simulated total exceeds the candidate stock.

Text, a sequence of characters, written in single, double or triple quotes.

Folds that keep each class's share.

Separate random draws within predeclared groups.

A split that keeps each class's share the same in every part.

A predeclared, nonoverlapping group used for selection; here United Kingdom, Germany, France and Other.

multiplying the output by a

Example: 2x²

every chord lies strictly above the curve; at most one minimiser

Example: x² (one), log(1 + eˣ) (none)

Text, written in quotes: "Hello, world!". Its type is str.

the total area of thin rectangles under a curve

Example: 0.46875 with four right-edge strips

A class defined from a base class: RatingError is a subclass of ValueError.

Records selected by one declared value of another variable.

The same measure split by a declared category, such as recorded country.

a small index written below and to the right

Example: x₂₁ = 0

A contiguous sequence of characters within a string.

Running many settings on little data and giving more data to the best.

One number computed from a whole data set, such as a correlation; it can not show the shape behind it.

add an expression over an index

Example: Σᵢ₌₁³ i² = 14

a small mark above and to the right

Example: x², or x⁽¹⁾ for row 1

Learning from examples whose labels are known.

A released value used with survey guidance to represent the target population in an analysis.

A summary distorted because only cases that passed a selection step were recorded: the simulated approvals default at 13.8%, the simulated rejects at 50.3%.

A = UΣVᵀ: rotate, stretch along the axes, rotate

Example: every matrix

equal to its own transpose; real eigenvalues, perpendicular eigenvectors

Example: XᵀX