WBCHSE HS Artificial Intelligence & Data Science flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

Turing Test

A test proposed by Alan Turing in 1950: a machine passes if a human interrogator cannot reliably distinguish its text responses from a human's.

Intelligent agent

An entity that perceives its environment through sensors and acts through actuators; a rational agent chooses actions expected to maximize its performance measure.

PEAS description

A framework for specifying an agent by its Performance measure, Environment, Actuators and Sensors; used to design and compare agent architectures.

Python list mutability

Lists are mutable sequences; indexing starts at zero, slicing excludes the stop index, and assigning one list to another shares the same object reference.

NumPy broadcasting

NumPy broadcasting lets arrays with compatible shapes be combined without copying data; dimensions are compared from the right and size-one dimensions stretch.

pandas loc versus iloc

loc selects by row and column labels, while iloc selects by integer positions; both accept slices and boolean masks for filtering DataFrame rows.

Function default arguments

Default argument values are evaluated once when a function is defined, so mutable defaults such as lists can retain changes across calls and should be avoided.

Missing value imputation

Replacing missing entries with estimated values such as the mean, median, mode or a model prediction; the method must be fitted on training data only.

IQR outlier rule

Values below Q1 minus 1.5 times the interquartile range or above Q3 plus 1.5 times the IQR are often flagged as outliers.

Data leakage

Using information unavailable at prediction time, such as test-set statistics or the target, during training; it produces optimistic evaluation and poor deployment performance.

Min-max normalization

Scales a numeric feature to a fixed range, usually zero to one, using (x minus minimum) divided by (maximum minus minimum); sensitive to outliers.

Standard deviation

The square root of variance; it measures the typical distance of observations from the mean in the same units as the data.

Central limit theorem

For sufficiently large sample sizes, the sampling distribution of the sample mean is approximately normal, regardless of the population distribution's shape.

Correlation coefficient

A number between minus one and plus one measuring the strength and direction of a linear relationship; zero indicates no linear association, not necessarily independence.

Conditional probability

The probability of event A given event B, written P(A|B) and computed as P(A and B) divided by P(B), provided P(B) is greater than zero.

Supervised learning

A machine-learning setting where a model learns a mapping from input features to known labeled targets, such as class labels or continuous values.

Overfitting

A model fits training data too closely, including noise, so training performance is high but generalization to new data is poor.

Precision and recall

Precision is true positives divided by predicted positives; recall is true positives divided by actual positives. F1 is their harmonic mean.

k-means clustering

An unsupervised algorithm that partitions data into k clusters by repeatedly assigning points to the nearest centroid and recomputing centroids as cluster means.

Perceptron

A linear classifier that computes a weighted sum of inputs plus bias and applies a step function; it cannot solve non-linearly separable problems such as XOR.

Backpropagation

An algorithm that applies the chain rule to compute gradients of the loss with respect to network weights, enabling gradient-based updates in multilayer neural networks.

ReLU activation

A nonlinear activation defined as max(0, z); it outputs zero for negative inputs and the input itself for positive values, reducing vanishing gradients.

Bag-of-words

A text representation that counts token occurrences per document and ignores word order, producing a sparse vector for each document.

TF-IDF

A weighting scheme multiplying term frequency in a document by inverse document frequency, so terms common across many documents receive lower weight.

Lemmatization

Reducing a word to its dictionary base form using vocabulary and grammatical rules, for example mapping 'running' to 'run'; more accurate than stemming.

Convolutional layer

A neural-network layer that slides learnable filters over an input to detect local patterns such as edges and textures while sharing weights across positions.

Max pooling

A downsampling operation that keeps the maximum value in each local window, reducing spatial dimensions and making features somewhat translation-invariant.

Data augmentation

Creating modified copies of training images through flips, rotations, crops or color changes to increase data variety and reduce overfitting.

Algorithmic bias

Systematic and unfair errors that disadvantage particular groups, often arising from biased training data, labels, sampling or proxy variables.

Data minimization

A privacy principle requiring collection and retention of only personal data necessary for a stated purpose, reducing breach and misuse risk.

Train-validation-test split

Partitioning data into training, validation and test sets; models fit on training data, hyperparameters tune on validation, and final performance is estimated on test data.

Baseline model

A simple reference predictor, such as the majority class or mean target, used to check whether a more complex model adds meaningful predictive value.