Difficulty: Intermediate

412 terms

Credit Scoring Intermediate

Predicting borrower default risk.

Cross-Attention Intermediate

Attention between different modalities.

Cross-Entropy Intermediate

Measures divergence between true and predicted probability distributions.

Cross-Validation Intermediate

A robust evaluation technique that trains/evaluates across multiple splits to estimate performance variability.

Curriculum Learning Intermediate

Ordering training samples from easier to harder to improve convergence or generalization.

Data Augmentation Intermediate

Expanding training data via transformations (flips, noise, paraphrases) to improve robustness.

Data Drift Intermediate

Shift in feature distribution over time.

Data Governance Intermediate

Processes and controls for data quality, access, lineage, retention, and compliance across the AI lifecycle.

Data Labeling Intermediate

Human or automated process of assigning targets; quality, consistency, and guidelines matter heavily.

Data Leakage Intermediate

When information from evaluation data improperly influences training, inflating reported performance.

Data Lineage Intermediate

Tracking where data came from and how it was transformed; key for debugging and compliance.

Data Poisoning Intermediate

Maliciously inserting or altering training data to implant backdoors or degrade performance.

Data Protection Impact Assessment Intermediate

Privacy risk analysis under GDPR-like laws.

Data Scaling Intermediate

Increasing performance via more data.

Dataset Intermediate

A structured collection of examples used to train/evaluate models; quality, bias, and coverage often dominate outcomes.

Dataset Shift Intermediate

Differences between training and deployed patient populations.

Datasheet for Datasets Intermediate

Structured dataset documentation covering collection, composition, recommended uses, biases, and maintenance.

Deep Learning Intermediate

A branch of ML using multi-layer neural networks to learn hierarchical representations, often excelling in vision, speech, and language.

Depth vs Width Intermediate

Tradeoffs between many layers vs many neurons per layer.

Differential Privacy Intermediate

A formal privacy framework ensuring outputs do not reveal much about any single individual’s data contribution.

Differential Progress Intermediate

Accelerating safety relative to capabilities.

Distillation Intermediate

Training a smaller “student” model to mimic a larger “teacher,” often improving efficiency while retaining performance.

Distribution Shift Intermediate

Train/test environment mismatch.

Domain Shift Intermediate

A mismatch between training and deployment data distributions that can degrade model performance.

DPO Intermediate

A preference-based training method optimizing policies directly from pairwise comparisons without explicit RL loops.

Dropout Intermediate

Randomly zeroing activations during training to reduce co-adaptation and overfitting.

Dual Problem Intermediate

Alternative formulation providing bounds.

Due Process Intermediate

Legal right to fair treatment.

E-Discovery Intermediate

AI-assisted review of legal documents.

Early Stopping Intermediate

Halting training when validation performance stops improving to reduce overfitting.