Advanced Concepts of Modeling in AI

Vocabulary flashcards covering the definitions of AI, ML, and DL, common data terminologies, modeling approaches, and the structural components of neural networks.

Artificial Intelligence (AI)

A technology that empowers machines to perform tasks traditionally requiring human intellect, such as reasoning, problem-solving, and decision-making.

Machine Learning (ML)

A specialized subset of AI that harnesses statistical techniques to enable machines to learn from experience and refine their performance over time without being explicitly programmed.

Deep Learning (DL)

A sophisticated subset of Machine Learning that uses multi-layered neural networks to process large amounts of data and solve complex problems like image and speech recognition.

Input (Raw Data)

The unprocessed data collected from multiple sources that serves as the foundation for a machine learning model to learn patterns.

ML Model

The processing unit that learns patterns and relationships from input data to generate intelligent predictions and actionable insights.

Segmentation

The process of grouping similar items or customers in a data-driven model to provide better targeting and personalization.

Classification

The task of assigning categories or labels to data for automated decision-making and efficiency.

Recommendation Systems

A real-world application of ML that analyses user preferences and browsing history to suggest personalized products or content.

Anomaly Detection

An ML application used to identify unusual patterns in data, such as identifying irregular heart rates or blood pressure in a medical context.

Features

The columns of a data table that describe the characteristics of a dataset, such as name, colour, and size in a fruit dataset.

Labels

Tags attached to data that identify its meaning, such as the name of a fruit given its features; data can be labelled or unlabelled.

Training Data

A collection of examples provided to an AI model to analyse and learn from, similar to a teacher using solved examples to teach a student.

Testing Data

A set of data used without labels to check the accuracy of a model's answers, similar to a class test.

Rule-Based Approach

A modeling method where patterns are defined by the developer through static if-then instructions, meaning the machine cannot learn from mistakes.

Learning-Based Approach

An adaptive modeling approach where the machine discovers patterns on its own by studying examples or receiving feedback.

Supervised Learning

A type of ML where the dataset is labelled and the model is trained under the guidance of defined pairs of features and labels.

Regression

A sub-type of supervised learning where the model predicts a continuous value, such as price, temperature, or age.

Unsupervised Learning

A type of ML where the model works on unlabelled data to discover relationships, patterns, and trends without human guidance.

Clustering

A sub-type of unsupervised learning where the model finds similarities between items and groups them without classes being defined in advance.

Association

A sub-type of unsupervised learning that finds interesting relationships between variables, such as identifying items customers are likely to buy together.

Reinforcement Learning

A type of machine learning where a system learns to make a series of decisions through trial and error to maximise a reward mechanism.

Artificial Neural Network (ANN)

A deep learning model composed of interconnected layers of neurons that can automatically extract features from large datasets and provide probability values for inputs.

Convolutional Neural Network (CNN)

A specific deep-learning algorithm built for processing images by assigning learnable weights and biases to different features.

Input Layer

The first layer of a neural network that acquires data and feeds it into the system without performing any processing.

Hidden Layers

The processing layers of a neural network where nodes run algorithms using weights, biases, and activation functions.

Output Layer

The final layer of a neural network that presents the processed result to the user without doing any processing itself.

Perceptron

The simplest building block of a neural network; a single artificial neuron that judge inputs by importance to produce a single yes-or-no decision.

Weights

Values in a perceptron or neural network that represent the importance of a specific input toward the final decision.

Bias

A value added to the weighted sum of inputs in a perceptron to shift the result before it is compared against a threshold.

Threshold

The value used in a perceptron calculation where if the total weighted sum is greater than or equal to this limit, the output is $1$, otherwise it is $0$.