Glossary
Quick reference for key terms used throughout this knowledge base. For detailed explanations, see Concepts.
Sections
- AI & Machine Learning — Core AI/ML/DL terms and GenAI/LLM concepts
- Metrics — Evaluation metrics
- Mechanical — Engineering terms
- Software Engineering — APIs, backend, and dev tooling
- Project Management — PM structures and methods
Using the Glossary
The glossary provides concise definitions of key terms. Each term links to related concepts for deeper learning.
Non-linear function applied to neuron output that enables networks to learn complex patterns.
- ReLU: max(0, x) - most common
- Sigmoid: 1/(1+e^-x) - outputs 0 to 1
- Tanh: (e^x - e^-x)/(e^x + e^-x) - outputs -1 to 1
- GELU: Gaussian Error Linear Unit - used in Transformers
An LLM-based system that autonomously decides which actions to take (tool calls, reasoning steps) to complete a goal, rather than responding in a single pass.
- Maintains a loop: observe → reason → act
- Can call tools, read/write files, browse, execute code
- Reliability depends on the model's instruction-following quality
Early deep CNN that revolutionized computer vision (2012).
- 8 layers, ReLU activation
- Dropout for regularization
- Won ImageNet competition
Identifying unusual or rare patterns in data.
- One-class SVM
- Autoencoder reconstruction error
- Industrial defect detection
Machine vision system that automatically inspects parts or PCBs for surface defects, dimensional deviations, or missing features.
- Replaces or augments manual visual inspection
- Uses cameras, structured light, or laser profiling
- Output: pass/fail classification + defect location
- The application context for industrial CV projects (PatchCore, YOLOv8 segmentation)
A server that acts as the single entry point for client requests, routing them to appropriate backend services and handling cross-cutting concerns.
- Responsibilities: authentication, rate limiting, logging, request routing
- Examples: Kong, AWS API Gateway, NGINX, custom FastAPI gateway
- Decouples clients from internal service topology
Mechanism that allows model to focus on relevant parts of input.
- Self-attention: relationships within input
- Cross-attention: relationships between different inputs
- Key-query-value (QKV) computation
Area Under the Receiver Operating Characteristic curve. Measures classifier performance across all decision thresholds.
- ROC curve: plots True Positive Rate vs False Positive Rate at each threshold
- AUC = 1.0: perfect classifier; AUC = 0.5: random
- Threshold-independent — useful when the operating point isn't fixed
- Used in PatchCore evaluation (image-level AUROC)
Two related but distinct security concepts:
- Authentication (authn): verifying who the caller is (e.g., API key, JWT, OAuth)
- Authorization (authz): verifying what the caller is allowed to do (e.g., admin vs. read-only)
You must authenticate before you can be authorized.
Algorithm for training neural networks by computing gradients of loss with respect to weights.
- Forward pass: compute output
- Backward pass: compute gradients
- Update weights using gradient descent
Technique that normalizes layer inputs to have zero mean and unit variance.
- Stabilizes training
- Allows higher learning rates
- Acts as regularizer
Number of samples processed before updating weights.
- Small: noisy gradient, better generalization
- Large: stable gradient, faster training
- GPU memory constraint
| Low Variance | High Variance | |
|---|---|---|
| High Bias | Underfitting | Good |
| Low Bias | Overfitting | Good |
- Bias: Error from overly simplistic assumptions
- Variance: Error from sensitivity to training data
- Goal: minimize total error = bias² + variance + irreducible error
Elastic instability of slender members under compressive load. Sudden lateral deflection at critical load.
- Euler buckling: P_cr = π²EI/(KL)²
- Governs design of columns and slender struts
- Failure can occur below material yield strength
- slenderness ratio (KL/r) determines behavior
Practices that automate building, testing, and deploying software.
- CI: every code change is automatically built and tested
- CD: validated changes are automatically deployed to staging or production
- Common tools: GitHub Actions, GitLab CI, Jenkins
- Reduces integration risk and speeds up delivery cycles
Predicting a categorical label for input data.
- Binary: two classes
- Multi-class: more than two classes
- Output: probability distribution over classes
Deep learning architecture designed for processing grid-like data (images).
- Convolution: Local pattern detection with shared weights
- Feature maps: Activation patterns for learned features
- Pooling: Downsampling for spatial invariance
- Early layers learn edges, deeper layers learn objects
A table showing the counts of TP, FP, FN, TN predictions, giving a complete picture of classifier behavior.
| Predicted Positive | Predicted Negative | |
|---|---|---|
| Actual Positive | TP | FN |
| Actual Negative | FP | TN |
- Foundation from which Precision, Recall, F1, and Accuracy are derived
- Reveals which specific error types the model makes
Packaging an application and its dependencies into a portable, isolated unit called a container.
- Containers share the host OS kernel (lighter than VMs)
Dockerfile: recipe for building an imagedocker-compose: define and run multi-container applications- Ensures consistent environments from development to production
The maximum number of tokens an LLM can process in a single forward pass (input + output combined).
- Determines how much prior conversation or document text can be included
- Typical range: 8K–200K tokens depending on model
- Exceeding it requires chunking, summarization, or retrieval
Operation that slides a kernel over input to produce feature maps.
- Kernel: small learned filter (e.g., 3x3)
- Stride: step size of sliding
- Padding: border handling
Time-dependent deformation under sustained load at elevated temperature.
- Significant above ~0.4 × T_melt (in Kelvin)
- Three stages: primary (decelerating), secondary (steady), tertiary (accelerating to failure)
- Critical in: turbine blades, nuclear reactors, high-T applications
- Design based on stress to produce 1% strain in 100,000 hours
Training technique that splits data into k folds, training k times and validating on different fold each time.
- k-fold CV, leave-one-out, stratified CV
- Provides more reliable performance estimate
- Helps detect overfitting
Creating modified versions of training data to increase diversity.
- Random crop, flip, rotation
- Color jittering
- Cutout, mixup, cutmix
- Improves generalization
Hierarchical structure that makes decisions by splitting data based on feature values.
- Interpretable, handles both numerical and categorical data
- Prone to overfitting (mitigated by pruning, ensembling)
- Split criteria: Gini impurity, information gain
Each layer connects to every other layer in a feedforward manner.
- Feature reuse
- Reduces parameters
- Strong gradient flow
A design pattern where a component receives its dependencies from the outside rather than creating them internally.
- Decouples components for easier testing and replacement
- Enables mock objects in unit tests
- Common in frameworks: FastAPI (
Depends), Angular, Spring
Engineering practice of designing parts to be easier, cheaper, and more reliable to manufacture.
- Minimize number of parts, reduce tight tolerances where possible
- Choose features compatible with available processes (drilling vs. milling, etc.)
- Evaluate early — changes are cheap at design stage, expensive in production
- Closely linked to tolerance stack-up and material selection
Generative models that learn to reverse a gradual noising process.
- Forward process: Adds noise to data
- Reverse process: Removes noise to generate
- Denoising Diffusion Probabilistic Models (DDPM)
Reducing feature count while preserving structure.
- PCA: Principal Component Analysis - orthogonal projections
- t-SNE: Non-linear embedding for visualization
- UMAP: Preserves both local and global structure
Regularization technique that randomly sets a fraction of neuron outputs to zero during training.
- Prevents co-adaptation of neurons
- Reduces overfitting
- Typically 0.1-0.5 dropout rate
Architecture that uniformly scales depth, width, and resolution.
- Compound scaling
- State-of-the-art efficiency
- Balanced model design
A dense vector representation of text (or other data) that captures semantic meaning.
- Similar meaning → vectors close in space
- Used in semantic search, RAG, and clustering
- Produced by embedding models (e.g.,
text-embedding-ada-002,nomic-embed-text)
The harmonic mean of Precision and Recall. Useful when you need a single metric that balances both.
- F1 = 2 × (Precision × Recall) / (Precision + Recall)
- Ranges from 0 (worst) to 1 (best)
- Preferred over accuracy when classes are imbalanced
- Penalizes extreme imbalance between precision and recall
The ratio of a material's strength to the applied stress, providing a design margin.
- FoS = σ_allow / σ_applied
- Based on yield (prevent plastic deformation) or ultimate (prevent fracture)
- Typical values: 1.5–5.0 depending on application
- Higher for safety-critical or uncertain loading
Quality inspection terms mapping directly to classification errors:
- False Reject (Type I error) — A good part is incorrectly rejected
- Wastes material and production time
- Maps to False Positive in defect detection
- False Accept (Type II error) — A defective part is incorrectly accepted
- Escapes to customer, potential safety/quality issue
- Maps to False Negative in defect detection
In manufacturing, false accepts are almost always worse than false rejects — a missed defect reaching the customer is far more costly than scrapping a good part.
| Metric | Detects | Reduces |
|---|---|---|
| High Precision | Few false rejects | Scrap waste |
| High Recall | Few false accepts | Customer escapes |
Failure under cyclic loading at stresses below yield strength. Progresses through crack initiation, propagation, and sudden fracture.
- Characterized by S-N curves (stress vs. cycles)
- Fatigue limit (steel) ≈ 0.5 × σ_uts
- Aluminum has no true fatigue limit
- Affected by: surface finish, mean stress, environment
Numerical simulation technique that divides a model into small elements to approximate stress, heat, or other physical quantities.
- Used to predict failure, optimize geometry, and replace physical prototypes
- Common tools: ANSYS, Abaqus, SolidWorks Simulation
- Results are only as valid as mesh quality, material data, and boundary conditions
Creating new input features from raw data to improve model performance.
- Domain knowledge is crucial
- Transformations: log, polynomial, encoding
- Interaction features
Choosing the most relevant features for the model.
- Filter methods (correlation, chi-square)
- Wrapper methods (forward, backward selection)
- Embedded methods (Lasso, tree-based importance)
Continuing training of a pre-trained model on a smaller, task-specific dataset to adapt its behavior.
- Updates model weights (unlike prompt engineering)
- Requires labeled data and compute
- Risk of catastrophic forgetting on original capabilities
A structured, proactive method for identifying potential failure modes in a design or process and assessing their impact.
| Column | Description |
|---|---|
| Failure Mode | How could this fail? |
| Effect | What happens when it fails? |
| Severity (S) | How bad is the effect? (1–10) |
| Occurrence (O) | How likely is failure? (1–10) |
| Detection (D) | How likely is it caught? (1–10) |
| RPN | Risk Priority Number = S × O × D |
- High RPN items are prioritized for redesign or additional inspection
- Required in IATF 16949 (automotive) and ISO 13485 (medical devices)
Two networks competing: generator creates fake samples, discriminator distinguishes real from fake.
- Generator: Creates synthetic samples
- Discriminator: Classifies real vs fake
- Training is a minimax game
A standardized system (ASME Y14.5) for defining part geometry and allowable variation using symbolic language on engineering drawings.
- Replaces ± coordinate tolerances with feature-based controls
- Key controls: flatness, perpendicularity, true position, concentricity, runout
- Ensures functional fit and interchangeability regardless of manufacturer
- Critical for communicating design intent to machining and inspection
Inception module architecture with parallel pathways at different scales.
- Inception modules: multiple filter sizes in parallel
- 22 layers deep
- Parameter efficient
Sequential ensemble method that builds trees to correct errors of previous trees.
- XGBoost, LightGBM, CatBoost are popular implementations
- Excellent performance on structured data
- Requires careful hyperparameter tuning
Features designed by domain experts.
- HOG (Histogram of Oriented Gradients)
- SIFT (Scale-Invariant Feature Transform)
- LBP (Local Binary Patterns)
Resistance to localized plastic deformation. Often correlated with strength.
| Test | Scale |
|---|---|
| Brinell (HB) | 500 kg load |
| Rockwell (HRC/HRB) | Scale-specific |
| Vickers (HV) | 1-1200 kg |
- Higher hardness generally means higher strength but lower ductility
Setting initial weights before training.
- Xavier/Glorot: for tanh, sigmoid
- He initialization: for ReLU
- Pre-training can help
Measures the overlap between a predicted bounding box (or mask) and the ground truth.
- IoU = Area of Overlap / Area of Union
- Ranges from 0 (no overlap) to 1 (perfect)
- Threshold (e.g., 0.5) determines what counts as a correct detection
- Also used in segmentation as pixel-level overlap metric
Instance-based learning that classifies based on k closest training examples.
- Simple, no training phase
- Sensitive to noise and high dimensionality
- Lazy learning - stores all training data
Small learnable filter in convolution operations.
- Detects specific patterns (edges, textures)
- Shared across spatial locations
- Stacking creates hierarchical features
A neural network (typically Transformer-based) trained on large text corpora to predict and generate language. Scales to billions of parameters.
- Examples: GPT-4, Claude, Llama, Mistral
- Capabilities: text generation, reasoning, code, summarization
- Deployed via API (cloud) or locally (Ollama, llama.cpp)
Controls how much model weights change during each update.
- Too high: unstable training, divergence
- Too low: slow convergence
- Learning rate schedules: warmup, decay
Predicting continuous values using a linear function: y = w^T x + b
- Minimizes Mean Squared Error (MSE)
- Highly interpretable baseline
- Assumes linear relationship between features and target
Binary classification using a sigmoid function to output probabilities: P(y=1|x) = sigmoid(w^T x + b)
- Output values between 0 and 1
- Decision boundary is linear
- Foundation for neural networks
Measures difference between predicted and actual values.
- MSE: Mean Squared Error - regression
- Cross-Entropy: Classification
- Binary Cross-Entropy: Binary classification
Advanced RNNs that handle long-term dependencies through gating mechanisms.
- LSTM: Long Short-Term Memory - input, forget, output gates
- GRU: Gated Recurrent Unit - simpler, fewer gates
- Solves vanishing gradient problem
The primary metric for object detection benchmarks. Averages the Average Precision (AP) across all object classes.
- AP: area under the Precision-Recall curve for one class
- mAP@0.5: IoU threshold of 0.5 to count a detection as correct
- mAP@0.5:0.95: averaged over IoU thresholds 0.50–0.95 (stricter, used in COCO)
- Standard for YOLO, Faster R-CNN, and similar detectors
Fully connected feedforward neural network with multiple layers.
- Each neuron connects to all neurons in next layer
- No spatial awareness
- Foundation for all deep learning
Efficient architecture designed for mobile/edge devices.
- Depthwise separable convolutions
- Very few parameters
- Trade-off: accuracy vs efficiency
A measure of material stiffness — the ratio of stress to strain in the linear elastic region.
- Symbol: E (GPa)
- Slope of stress-strain curve in elastic region
- Steel ≈ 200 GPa, Aluminum ≈ 70 GPa
- Higher E = stiffer material
Computing system inspired by biological neural networks, consisting of interconnected nodes (neurons) that process information through weighted connections.
- Neuron: Basic unit that computes weighted sum + activation
- Layer: Group of neurons
- Deep: Multiple hidden layers
Finding and classifying objects in images with bounding boxes.
- Single-stage: YOLO (fast)
- Two-stage: Faster R-CNN (accurate)
- Output: bounding boxes + class labels
Algorithm that updates model parameters to minimize loss.
- SGD: Stochastic Gradient Descent
- Adam: Adaptive moment estimation - most popular
- AdamW: Adam with weight decay
A library that maps between database tables and language objects, letting developers query the database using code instead of raw SQL.
- Examples: SQLAlchemy (Python), Prisma (TypeScript), Django ORM
- Handles query building, type mapping, and connection management
- Trade-off: convenience vs. control over generated SQL
Definition: The end result or consequence of a process or project.
Key idea: Focuses on benefits and value, not just deliverables.
Notes to remember
- Includes outputs and artifacts
- But mainly about what value or benefit is achieved
Model learns training data too well, including noise, failing to generalize.
Signs: High train accuracy, low validation accuracy
Remedies: More data, regularization, dropout, early stopping, simpler model
The ratio of lateral strain to axial strain when a material is loaded axially.
- Symbol: ν (nu)
- Typically 0.25–0.35 for metals
- 0.5 for incompressible materials (rubber)
- Negative values exist for auxetic materials
Downsampling operation that reduces spatial dimensions.
- Max pooling: takes maximum value in region
- Average pooling: takes average value
- Provides translation invariance
Definition: A collection of projects, programs, subsidiary portfolios, and operations managed together to achieve strategic objectives.
Key idea: Portfolio = Strategy level
Memory tip
Portfolio ensures the organization invests in the right work.
See Metrics → Precision / Recall for the full definition and manufacturing context.
Two complementary metrics for evaluating classification performance, especially when classes are imbalanced.
- Precision = TP / (TP + FP) — "Of all positive predictions, how many were correct?"
- Recall = TP / (TP + FN) — "Of all actual positives, how many did we find?"
| High Precision | Low Precision | |
|---|---|---|
| High Recall | Ideal | Many false positives |
| Low Recall | Missing real positives | Poor model |
Trade-off: Increasing recall usually decreases precision and vice versa. The right balance depends on the application cost of false positives vs false negatives.
Definition: A quantifiable artifact produced by a project that can be:
- a final product, or
- a component of another product
Examples
- Software application
- Car engine
- Mobile phone component
Definition: A group of related projects managed together to obtain benefits not available if managed separately.
Key idea Program = Coordination for extra benefits
Example Airport expansion program:
- Runway project
- Terminal project
- Parking system project
Definition: A temporary endeavor undertaken to create a unique product, service, or result.
Key characteristics
- Temporary → has a start and end
- Unique → not routine work
Notes
- Can exist alone
- Or be part of a program or portfolio
Definition: The application of knowledge, skills, tools, and techniques to project activities to meet project requirements.
Key idea Project management guides work to deliver intended outcomes.
Approaches
- Predictive (Waterfall)
- Hybrid
- Adaptive (Agile)
Definition: The person assigned by the performing organization to lead the project team and achieve project objectives.
Key responsibilities
- Facilitate team work
- Manage project processes
- Ensure outcomes are delivered
Definition: A group of individuals performing the work of the project to achieve its objectives.
Includes
- Specialists
- Engineers
- Developers
- Analysts
- Any role needed for project work
Designing inputs to an LLM to elicit accurate, useful, or structured outputs — without changing model weights.
- Zero-shot: no examples provided
- Few-shot: a few examples in the prompt
- Chain-of-thought: ask the model to reason step by step
- Directly relevant to agentic system design
Ensemble of decision trees trained on random subsets of data and features.
- Reduces overfitting through averaging
- Handles missing values, high-dimensional data
- Robust to outliers
A prompting pattern for agents that interleaves chain-of-thought reasoning with action execution.
- Thought: model reasons about what to do next
- Action: model calls a tool or takes a step
- Observation: result is fed back into context
- Used in the Coding Agent project
Region of input that influences a particular neuron's output.
- Shallow layers: small receptive field (local patterns)
- Deep layers: large receptive field (global structure)
- Stacking 3x3 convolutions is more efficient than one large kernel
Techniques to prevent overfitting.
- L1/L2 weight decay
- Dropout
- Data augmentation
- Early stopping
CNN with residual connections (skip connections) enabling very deep networks.
- Residual block: Output = F(x) + x
- Allows gradient flow in deep networks
- Enables training of 100+ layer networks
An architectural style for networked applications where clients communicate with servers over HTTP using standard methods.
- GET: retrieve a resource
- POST: create a resource
- PUT / PATCH: update a resource
- DELETE: remove a resource
- Stateless — each request contains all needed information
- Resources identified by URLs; responses typically JSON
Pattern that augments LLM responses by retrieving relevant documents from an external knowledge base at query time.
- Reduces hallucinations on factual queries
- Keeps knowledge base updatable without retraining
- Components: retriever (embedding + vector DB) + generator (LLM)
Network with loops allowing information to persist, processing sequential data.
- Processes one element at a time
- Maintains hidden state
- Suffers from vanishing gradients (long sequences)
Pixel-level classification in images.
- Semantic segmentation: class per pixel
- Instance segmentation: distinguish individual objects
- Panoptic segmentation: both
Learning from both labeled and unlabeled data. Leverages abundant unlabeled data alongside limited labeled data to improve model performance.
A lightweight, serverless, file-based relational database.
- No separate server process — the database is a single
.dbfile - Suitable for local apps, prototypes, and low-concurrency services
- Used in the Ollama Gateway for API key and usage tracking
- Not suitable for high-concurrency production write workloads
Localized increase in stress near geometric discontinuities (holes, notches, keyways).
- kt = theoretical stress concentration factor
- Causes: sudden cross-section changes, sharp corners
- Mitigate with: fillets, gradual transitions
- Critical under cyclic loading (fatigue)
Learning from labeled data where inputs are mapped to known outputs. The model learns a function that maps inputs to correct outputs through training on example pairs.
- Classification: Predicting discrete categories/labels
- Regression: Predicting continuous values
- Training data: input-output pairs (x, y)
- Goal: minimize prediction error on unseen data
A quantitative measure of the texture/finish of a machined surface, typically expressed as Ra (arithmetic average roughness).
- Ra — average deviation from the mean line (µm)
- Lower Ra = smoother surface
- Affected by: tool condition, feed rate, cutting speed, material
| Finish Type | Typical Ra (µm) |
|---|---|
| Fine grinding | 0.1 – 0.4 |
| Turning | 0.8 – 3.2 |
| Milling | 1.6 – 6.3 |
| As-cast | 6.3 – 25 |
Finds the maximum-margin hyperplane between classes.
- Support vectors define the decision boundary
- Kernel trick handles non-linear boundaries
- Effective in high-dimensional spaces
Hierarchical Vision Transformer with shifted window attention.
- Linear complexity
- Local and global attention
- State-of-the-art for many vision tasks
Definition: A collection of strategic business activities used to build, sustain, or advance an organization.
Includes
- Portfolios
- Programs
- Projects
- Products
- Operations
Key idea Everything works together to deliver value.
Stress induced in a constrained material due to temperature change.
- σ = EαΔT (for unconstrained thermal expansion)
- α = coefficient of thermal expansion
- Can cause buckling or fracture if constrained
- Critical in assemblies with mismatched thermal expansion
The cumulative effect of individual part tolerances in an assembly. When multiple parts are assembled, their tolerances combine — potentially causing fit or function issues.
Two analysis methods:
- Worst-case: Sum of all tolerances (conservative, guarantees 100% assembly)
- Statistical (RSS): Root sum of squares — assumes normal distribution, allows tighter design with acceptable risk
Tolerance stack-up analysis is critical for multi-part assemblies where clearance or interference must be controlled.
A capability where an LLM outputs a structured request to invoke an external function, rather than generating free text.
- Model outputs: function name + arguments (JSON)
- Host code executes the function and returns the result
- Enables agents to interact with APIs, filesystems, databases
Ability to absorb energy before fracture. Combination of strength and ductility.
- Impact toughness: measured by Charpy/Izod tests
- Fracture toughness (K_IC): resistance to crack propagation
- Higher toughness = more ductile, less brittle
Architecture based on self-attention, processing sequences in parallel.
- Self-Attention: Computes relationships between all positions
- Multi-head Attention: Multiple attention patterns in parallel
- Positional Encoding: Adds sequence order information
- Foundation of modern NLP and vision models
The maximum stress a material can withstand while being stretched or pulled before necking and fracture.
- Symbol: σ_uts or σ_u (MPa)
- Found at peak of engineering stress-strain curve
- Used for calculating factor of safety against fracture
- Note: different from true stress at failure
Model fails to capture patterns in training data.
Signs: Low accuracy on both train and validation
Remedies: More complex model, more features, more training time
Encoder-decoder architecture with skip connections for image segmentation.
- Symmetric encoder-decoder
- Skip connections preserve spatial information
- Used in medical imaging, semantic segmentation
Learning patterns from unlabeled data without explicit labels. The model discovers structure, patterns, or representations inherent in the data.
- Clustering: Grouping similar data points
- Dimensionality Reduction: Reducing feature complexity while preserving structure
- Density Estimation: Modeling data distribution
- No labeled outputs - model finds patterns autonomously
Definition: The worth, importance, or usefulness of something.
Important concept Different stakeholders see value differently.
Types of value
- Customer value → product features and usability
- Business value → financial benefit (benefits – cost)
- Societal value → benefits to society or environment
A database optimized for storing and querying high-dimensional vectors (embeddings) by similarity.
- Query type: "find the k nearest vectors"
- Examples: Chroma, Qdrant, Pinecone, pgvector
- Core infrastructure for RAG pipelines
Simple, uniform architecture with 3x3 convolutions throughout.
- 16-19 layers deep
- Proven that deeper networks work better
- Foundation for modern architectures
Applying Transformer architecture to image recognition.
- Splits images into patches
- Processes patches as sequence
- Requires more data than CNNs but scales better
Learning from coarse or incomplete labels. Uses inexact, imprecise, or noisy labels rather than exact annotations.
Learnable parameters in neural networks.
- Weights (w): Control strength of connections between neurons
- Bias (b): Offset added to weighted sum
- Learned through backpropagation
The stress at which a material begins to deform permanently (plastically). Below yield strength, deformation is elastic and reversible.
- Symbol: σ_y (MPa)
- Determined from stress-strain curve (0.2% offset method)
- Critical for structural design — parts must operate below yield to avoid permanent deformation
Single-stage object detection that predicts bounding boxes directly.
- Single forward pass
- Real-time detection
- Trade-off: speed vs accuracy