ASSEB Class 10 AI & Robotics flashcards

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

Narrow AI

An AI system designed for one specific task or a limited domain, such as face unlock or spam filtering, and unable to transfer its skill to unrelated problems.

Supervised learning

A learning approach where the model is trained on labelled input-output pairs and learns a mapping that can predict outputs for new, unseen inputs.

Reinforcement learning

A learning approach where an agent takes actions in an environment and improves its policy by maximizing cumulative reward received through trial and error.

Classification versus regression

Classification predicts a discrete class label such as spam or not spam; regression predicts a continuous numeric value such as temperature or price.

Ordinal data

Categorical data whose values have a meaningful order but unequal or undefined spacing between them, such as small, medium, and large shirt sizes.

Median robustness

The median is the middle value of sorted data and is less affected by extreme outliers than the mean, making it preferred for skewed distributions.

Data leakage

A modelling error where information unavailable at prediction time, such as test-set or target information, enters training and inflates reported performance.

Correlation coefficient r

A number between minus one and plus one describing the strength and direction of a linear relationship; a value near zero indicates no linear relationship.

Overfitting

A condition where a model fits training data very closely, including its noise, and therefore performs poorly on new data because it has high variance.

Precision and recall

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

Confusion matrix

A table of true positives, true negatives, false positives, and false negatives that shows which classes a classifier confuses and where its errors concentrate.

k-means clustering

An unsupervised algorithm that partitions data into k groups by repeatedly assigning points to the nearest centroid and recomputing centroids until stable.

Convolution operation

Sliding a small kernel of weights across an image grid and computing weighted sums, producing feature maps that respond to local patterns such as edges.

Semantic segmentation

A computer vision task that assigns a class label to every pixel of an image, without separating individual instances of the same class.

TF-IDF

A text weighting scheme that multiplies term frequency in a document by inverse document frequency across the corpus, downweighting words that appear everywhere.

Word embeddings

Dense numeric vectors representing words so that semantically similar words lie close together in vector space, enabling similarity and analogy operations.

Intersection over Union

A detection and segmentation metric equal to the overlap area of predicted and true regions divided by their combined area, ranging from zero to one.

Proprioceptive sensors

Sensors that measure the robot's own internal state, such as wheel encoders for rotation and inertial measurement units for orientation and acceleration.

Ultrasonic distance formula

Distance equals speed of sound multiplied by time of flight divided by two, because the pulse travels to the object and back to the sensor.

Degrees of freedom

The number of independent joints or motion variables a robot has, determining the range of positions and orientations its end effector can reach.

Inverse kinematics

The computation of joint angles needed to place a robot's end effector at a desired position and orientation; it may have multiple or no solutions.

PID controller

A feedback controller combining proportional response to current error, integral response to accumulated error, and derivative response to the rate of change.

Pulse width modulation

A digital technique that approximates an analog output by rapidly switching a pin on and off and varying the fraction of time it stays on.

H-bridge motor driver

A circuit that lets a microcontroller reverse a DC motor's direction and supply higher current, isolating motor power from sensitive logic electronics.

Servo motor

A motor with internal position feedback that rotates to and holds a commanded angle, typically controlled by a PWM pulse width.

ADC resolution

The number of discrete levels an analog-to-digital converter produces; a 10-bit ADC maps an input voltage range to integer values from zero to 1023.

I2C bus

A two-wire serial protocol using data and clock lines, where multiple addressed devices share the bus, suited to connecting several low-speed sensors.

Baseline model

A simple reference model, such as a majority-class predictor, whose score a more complex model must beat to justify its added complexity.

Reproducibility in projects

The practice of fixing random seeds and recording dataset versions, hyperparameters, and library versions so that reported results can be regenerated later.

Algorithmic bias

Systematically unfair model outcomes arising from biased training data, labelling, or proxy features, producing different error rates across groups.

Data minimization

A privacy principle limiting collection and retention to the personal data actually necessary for a stated purpose, reducing harm from breaches or misuse.

Fail-safe design

A safety approach where a system defaults to a safe state, such as stopping motors, when power, communication, or sensor input is lost.