AI glossary: terms starting with L

67 terms starting with L, each defined the way the lesson that teaches it uses it, with a link to every lesson where it appears.

L∞ (max) norm

max \ | vᵢ\ |

Example: 4

Taught in
L1 (Manhattan) norm

Σ \ | vᵢ\ |

Example: 7

Taught in
L1 norm

The sum of absolute values; its unit ball is a diamond.

Taught in
L1 penalty

The sum of absolute weights.

Taught in
L1 ratio

The share of the L1 penalty in that mix.

Taught in
L2 (Euclidean) norm

√(Σ vᵢ²), the straight-line length

Example: 5 for (3, −4)

Taught in
L2 norm

The square root of the sum of squares.

Taught in
L2 penalty

The sum of squared weights.

Taught in
label

The answer a program should give for an example.

Taught in
label (target)

The column the model predicts: mpg.

Taught in
label noise

Labels that are wrong.

Taught in
label vector

The labels, one per sample, written y.

Taught in
lag

A fixed number of time steps between paired observations.

Taught in
Lagrange multiplier

λ in ∇f = λ∇g at a constrained optimum

Example: λ = −2

Taught in
Lagrangian

L = f − λ(g − c); its stationary point gives the constrained optimum

Example: three equations for x, y, λ

Taught in
lambda

A function written as one expression, without a name: lambda price: round(price * 1.08, 2).

Taught in
lasso

Linear regression with a penalty on the absolute size of the weights.

Taught in
lasso path

Every weight plotted as α decreases.

Taught in
law of large numbers

With repeated independent draws from the same distribution, a running fraction tends toward that distribution's probability.

Taught in
law of total probability

The probability of the evidence is the sum over all disjoint paths that produce it: 108 flags per 1,000 messages, from spam and from real mail.

Taught in
lazy learner

A model that stores the training rows and does its work when it predicts.

Taught in
learning curve

Training and validation score against the number of training rows.

Taught in
learning rate

The hyperparameter that scales each step, η.

Taught in
learning-rate schedule

A rule that changes the learning rate during training.

Taught in
least squares

the x that makes the squared length of the residual smallest

Example: x̂ = (1.5, 0.5)

Taught in
leftover

The candidate stock exceeds the simulated total.

Taught in
leverage

How far a row's features are from the other rows', which sets how much it can turn the fit.

Taught in
library

Published code your program can import and use: NumPy, pandas, Matplotlib, scikit-learn, TensorFlow, PyTorch.

Taught in
likelihood

Probability of this evidence given a specified class: P(flagged given spam) = 90%.

Taught in
limit

the value a quantity approaches as h goes to zero

Example: 2 + h → 2

Taught in
linear

a straight-line graph, ax + b

Example: 2x + 1

Taught in
linear association

A straight line pattern between two numerical variables.

Taught in
linear classifier

predicts the class from the sign of w · x + b

Example: the ward line, 14 of 14

Taught in
linear combination

vectors, each multiplied by a number, then added

Example: 3·col₁ − 1·col₂

Taught in
linear independence

no column is a combination of the others

Example: (1, 1, 1) and (0, 1, 2)

Taught in
linear model

a prediction that is a straight-line function of its input

Example: price = 21.425 − 1.228 · age

Taught in
linear regression

Predicting a number with a straight line, or with a weighted sum of features.

Taught in
linear score

An intercept plus a coefficient times the feature, before the sigmoid.

Taught in

Checking the values one at a time from the start: O(n).

Taught in
Linear system

Equations represented as A @ w = b

Taught in
linear time

O(n): ten times the input, ten times the operations.

Taught in
linear transformation

moves the plane so grid lines stay straight, parallel and evenly spaced, with the origin fixed

Example: every matrix

Taught in
linter

A tool that reads code for mistakes, without running it: ruff check.

Taught in
list

Values in square brackets, in order, separated by commas: [8200, 10450, 6300]. Module 2 uses lists only in loops; Python 3.1 teaches them.

Taught in
list comprehension

A list built from an expression evaluated for each item of an iterable: [round(p * 1.08, 2) for p in prices], with an optional if at the end to filter.

Taught in
load factor

The number of keys divided by the number of buckets. A lookup's work grows with it.

Taught in
local derivative

the derivative of one node with respect to its own input

Example: ∂f/∂v = 2v = 10

Taught in
local minimum

lower than every nearby point

Example: 1.347, value −2.6186

Taught in
local variable

A name in a function's local scope; enclosing bindings can survive through a returned closure.

Taught in
lockfile

A file recording the exact version of every dependency: uv.lock.

Taught in
log likelihood

Sum of log probabilities assigned to observed outcomes.

Taught in
log loss

Mean negative log likelihood per observed row.

Taught in
log scale

an axis where equal distances are equal ratios

Example: 1, 0.1, 0.01 evenly spaced

Taught in
log transform

Applying a logarithm to each declared target value before fitting.

Taught in
log-odds (logit)

ln(p / (1 − p)); the sigmoid turns it back into a probability

Example: 1.3863 for p = 0.8

Taught in
log-sum-exp

log Σ e^(zᵢ), computed as m + log Σ e^(zᵢ − m) with m the largest zᵢ

Example: 1001.3133 for [1000, 1001]

Taught in
logarithm

log₂ x is the power two must be raised to to give x

Example: log₂ 4 = 2

Taught in
logarithmic time

O(log n): the operations grow by one each time n doubles. Binary search on 1,000,000 values makes at most 20 comparisons.

Taught in
logistic regression

Model that maps a feature-dependent linear score to a probability.

Taught in
long right tail

Relatively few observations occupy a wide high-value range.

Taught in
long-run average

The value a running average of repeated random draws tends toward: the expected value, by the law of large numbers.

Taught in
loop

Code that runs its body again and again: a while statement or a for statement, with its block.

Taught in
loop body

The indented block that repeats.

Taught in
loop else

A block written at the same level as a for or while, that runs when the loop ends without break.

Taught in
loop variable

The name a for loop binds to each item in turn: steps in for steps in week:. After the loop it retains the last value assigned to it; an empty loop does not bind it.

Taught in
loss curve

The loss plotted after every update.

Taught in
low-rank approximation

keep the k largest singular values; the best possible approximation of rank k

Example: digits, k = 21

Taught in