Glossary

Absolute Error is the absolute difference between a predicted value and the actual value. It measures how far off a prediction is without considering whether the error is positive or negative. It is commonly used in regression tasks to gauge the magnitude of prediction errors.

Accuracy is the proportion of correct predictions over the total number of predictions made. It is frequently used in classification problems to gauge how often the model gets the right label. However, it can be misleading when classes are highly imbalanced.

An activation function determines the output of a neuron given an input or set of inputs. It introduces non-linearity into neural networks, enabling them to learn complex patterns. Common examples include ReLU, Sigmoid, and Tanh.

Actor-Critic is a method in reinforcement learning that combines policy-based (actor) and value-based (critic) approaches. The actor decides on actions while the critic estimates the value function to guide the actor's updates. This approach often converges faster and more stably than purely policy- or value-based methods.

Adam (Adaptive Moment Estimation) is an optimization algorithm that combines ideas from RMSProp and momentum. It adapts the learning rate for each parameter, using running averages of both the gradients and their squares. Adam is popular due to its relatively fast convergence and ease of tuning.

Adversarial Examples are inputs crafted to fool an AI model into making incorrect predictions. They often involve small perturbations that are imperceptible to humans but significantly affect the model's output. Studying them helps improve model robustness and security against malicious attacks.

Aggregation in machine learning typically refers to combining multiple outputs, features, or predictions into a single representative value. This can occur at different stages, such as feature engineering (aggregating statistics) or ensemble methods (aggregating predictions). Proper aggregation can reduce noise and improve model performance.

An agent is an entity capable of perceiving its environment through sensors and acting upon it through actuators. In AI, agents typically strive to achieve specific goals or optimize a utility function. They can be software-based (like chatbots) or embodied (like robots).

Algorithmic Bias occurs when a machine learning system systematically and unfairly discriminates against certain individuals or groups. It can arise from biased training data, flawed assumptions, or imbalanced representations. Mitigating algorithmic bias is crucial for ethical and equitable AI deployment.

Alpha-Beta Pruning is a search algorithm that reduces the number of nodes evaluated in the minimax strategy for game-playing AI. It prunes branches that cannot possibly influence the final decision, increasing efficiency. This technique is especially common in two-player games like chess or checkers.