B (16 terms)
Backdoor / Trojan
Hidden behavior activated by specific triggers, causing targeted mispredictions or undesired outputs.
Intermediate
Batch Inference
Running predictions on large datasets periodically.
Intermediate
Batch Size
Number of samples per gradient update; impacts compute efficiency, generalization, and stability.
Intermediate
Bayesian Inference
Updating beliefs about parameters using observed evidence and prior distributions.
Intermediate
Beam Search
Search algorithm for generation that keeps top-k partial sequences; can improve likelihood but reduce diversity.
Intermediate
Behavior Cloning
Learning action mapping directly from demonstrations.
Advanced
Bellman Equation
Fundamental recursive relationship defining optimal value functions.
Intermediate
Benchmark
A dataset + metric suite for comparing models; can be gamed or misaligned with real-world goals.
Intermediate
Bias
Systematic differences in model outcomes across groups; arises from data, labels, and deployment context.
Intermediate
Bias Term
Systematic error introduced by simplifying assumptions in a learning algorithm.
Intermediate
Bias–Variance Tradeoff
A conceptual framework describing error as the sum of systematic error (bias) and sensitivity to data (variance).
Intermediate
Blackboard System
Agents communicate via shared state.
Advanced
Blue-Green Deployment
Maintaining two environments for instant rollback.
Intermediate
Boltzmann Machine
Probabilistic energy-based neural network with hidden variables.
Intermediate
Bottleneck Layer
A narrow hidden layer forcing compact representations.
Intermediate
Brier Score
A proper scoring rule measuring squared error of predicted probabilities for binary outcomes.
Intermediate