A
- Algorithm
- A set of instructions for solving a problem or performing a computation.
- Artificial Intelligence (AI)
- The field of computer science focused on creating systems that can perform tasks requiring human-like intelligence.
- Artificial General Intelligence (AGI)
- A hypothetical AI system capable of performing any intellectual task a human can do.
- Adversarial Attack
- A technique where small, intentional changes to input data cause an AI model to make errors.
- Agent
- An AI entity that perceives its environment and takes actions to achieve goals.
- Attention Mechanism
- A neural network component that helps models focus on relevant parts of input data, widely used in transformers.
B
- Backpropagation
- The process of training neural networks by adjusting weights based on errors.
- Bias (Algorithmic)
- Systematic unfairness in AI outputs, often caused by skewed training data.
- Big Data
- Extremely large datasets used to train AI models.
- Black Box Model
- An AI system whose decision-making process is not easily interpretable.
- Bayesian Network
- A probabilistic model representing variables and their conditional dependencies.
C
- Chatbot
- An AI system designed to simulate conversation with humans.
- Classifier
- A model that assigns categories to input data.
- Clustering
- An unsupervised learning technique for grouping similar data points.
- Computer Vision
- The AI field focused on enabling machines to interpret visual data.
- Concept Drift
- When the statistical properties of data change over time, reducing model accuracy.
- Convolutional Neural Network (CNN)
- A neural network architecture specialized for image processing.
D
- Data Mining
- Extracting useful patterns or insights from large datasets.
- Data Poisoning
- Malicious manipulation of training data to corrupt model performance.
- Decision Tree
- A simple model that makes predictions by following rules based on feature values.
- Deep Learning
- A subset of machine learning using multi-layered neural networks.
- Diffusion Models
- Generative AI models that create images or other data by iteratively removing noise.
E
- Edge AI
- Running AI models directly on devices (phones, IoT) rather than in the cloud.
- Embeddings
- Numerical representations of data (like words or images) that capture meaning and relationships.
- Ensemble Learning
- Combining multiple models to improve prediction accuracy.
- Explainable AI (XAI)
- Techniques to make AI decisions transparent and understandable.
- Evolutionary Algorithm
- An optimization method inspired by natural selection.
F
- Federated Learning
- Training AI models across decentralized devices while keeping data local.
- Feature Engineering
- Selecting and transforming input variables to improve model performance.
- Fine-Tuning
- Adapting a pre-trained model to a specific task or dataset.
- Foundation Model
- Large, general-purpose models (e.g., GPT, BERT) used as a base for multiple tasks.
G
- Generative AI
- AI that creates new content such as text, images, or music.
- Generative Adversarial Network (GAN)
- A model where two networks (generator and discriminator) compete to create realistic data.
- Gradient Descent
- An optimization algorithm used to minimize error in training neural networks.
- Graph Neural Network (GNN)
- A neural network designed to process graph-structured data.
H
- Hallucination (AI)
- When an AI system produces false or fabricated information with confidence.
- Heuristic
- A practical rule-of-thumb method for problem-solving, not guaranteed to be optimal.
- Hyperparameters
- Settings that control how a model is trained (e.g., learning rate, batch size).
I
- Inference
- Using a trained AI model to make predictions on new data.
- Interpretability
- The degree to which humans can understand how a model makes decisions.
- Intelligent Agent
- A system that perceives, decides, and acts in an environment to achieve goals.
J
- Jailbreak (AI)
- A method to bypass safety or ethical constraints built into an AI system.
- Jupyter Notebook
- An open-source tool used for coding, visualizations, and AI experiments.
K
- Knowledge Graph
- A structured representation of entities and their relationships.
- K-Nearest Neighbors (KNN)
- A simple algorithm that classifies data based on the closest examples in memory.
- Kernel Method
- Techniques in machine learning that transform data into higher dimensions for better separability.
L
- Large Language Model (LLM)
- An AI system trained on massive text datasets to generate human-like language.
- Latent Space
- A compressed, abstract representation of data inside a model.
- Learning Rate
- A hyperparameter that controls how much weights are updated during training.
- Logistic Regression
- A model used for binary classification tasks.
M
- Machine Learning (ML)
- A branch of AI where systems learn from data rather than being explicitly programmed.
- Markov Decision Process (MDP)
- A mathematical framework for decision-making in uncertain environments.
- Meta-Learning
- “Learning to learn,” where models adapt quickly to new tasks.
- Model Drift
- When a model’s performance degrades because real-world data changes.
N
- Natural Language Processing (NLP)
- AI focused on understanding and generating human language.
- Neural Network
- A system of interconnected nodes (neurons) inspired by the human brain.
- Normalization
- Adjusting data values to a consistent scale for training.
O
- Overfitting
- When a model learns the training data too well, including noise, and performs poorly on new data.
- Ontology
- A formal representation of concepts and relationships in a domain.
- Optimization
- The process of adjusting parameters to minimize error or maximize reward.
P
- Parameter
- Internal values in a model (like weights in a neural network) learned during training.
- Perceptron
- One of the earliest types of neural networks, capable of binary classification.
- Pretraining
- Initial training of a large model on broad data before fine-tuning for specific tasks.
- Prompt Engineering
- The practice of crafting inputs to guide AI outputs effectively.
Q
- Q-Learning
- A reinforcement learning algorithm that learns the value of actions in states.
- Quantum Machine Learning
- The use of quantum computing techniques to improve AI algorithms.
R
- Reinforcement Learning (RL)
- Training agents through rewards and penalties for actions.
- Regularization
- Techniques to prevent overfitting by constraining models.
- Retrieval-Augmented Generation (RAG)
- Combining information retrieval with generative models for more accurate results.
- Robotics
- The integration of AI into machines capable of physical action.
S
- Scalability
- The ability of a model or system to handle increasing amounts of data or computation.
- Semi-Supervised Learning
- Training models on a mix of labeled and unlabeled data.
- Speech Recognition
- AI systems that convert spoken language into text.
- Supervised Learning
- Training models using labeled datasets.
- Swarm Intelligence
- AI inspired by the collective behavior of decentralized systems (ants, bees, etc.).
T
- Tokenization
- Breaking down text into smaller units (tokens) for processing by models.
- Transformer
- A neural architecture that uses attention mechanisms, central to modern LLMs.
- Transfer Learning
- Reusing a pre-trained model on a new, related task.
- Turing Test
- A test of a machine’s ability to exhibit human-like intelligence in conversation.
U
- Unsupervised Learning
- Training models on unlabeled data to find hidden structures or patterns.
- Underfitting
- When a model is too simple to capture the complexity of data.
- Uncertainty Quantification
- Measuring confidence in AI predictions.
V
- Variational Autoencoder (VAE)
- A generative model used for creating new data from latent representations.
- Vector Database
- A database designed for storing and querying embeddings.
- Vision Transformer (ViT)
- A transformer architecture adapted for image recognition tasks.
W
- Weak AI
- AI designed for narrow tasks, unlike AGI which would cover general intelligence.
- Weight
- The numerical value in a neural network that determines the strength of connections between nodes.
- Word Embeddings
- Vector representations of words that capture semantic meaning.
X
- XAI (Explainable AI)
- Methods and techniques for making AI decisions understandable.
- XOR Problem
- A classic issue in neural networks that led to the development of multi-layer perceptrons.
Y
- YOLO (You Only Look Once)
- A real-time object detection system used in computer vision.
- Yield Curve Modeling (AI in Finance)
- The application of AI to predict shifts in financial markets.
Z
- Zero-Shot Learning
- The ability of a model to perform tasks without prior explicit training examples.
- Z-Score Normalization
- A statistical method for scaling data before feeding it into AI models.