AI Factory Glossary

garch, garch, time series models

Generalized AutoRegressive Conditional Heteroskedasticity models time-varying volatility in financial time series.

gas adsorption porosimetry, metrology

Use gas adsorption isotherms.

gas cabinet, manufacturing operations

Gas cabinets safely house compressed gas cylinders with monitoring and shutoff.

gas cabinet,facility

Enclosed ventilated cabinet for storing toxic or hazardous gas cylinders.

gas distribution, manufacturing operations

Gas distribution systems route specialty gases from storage to process tools.

gat multi-head, gat, graph neural networks

Multi-head attention in GAT computes multiple attention mechanisms in parallel stabilizing learning and improving expressiveness.

gat, gat, graph neural networks

Graph Attention Networks compute node representations by applying self-attention to neighborhood aggregation with learned attention coefficients.

gate dielectric interface, process integration

Gate dielectric interface quality affects mobility threshold voltage and reliability requiring careful engineering.

gate oxide,diffusion

Thin high-quality oxide under transistor gate critical for performance.

gate replacement,process

Remove dummy gate and fill with real gate.

gate stack, process integration

Gate stacks consist of gate dielectric and gate electrode layers whose materials thickness and interfaces determine transistor electrical characteristics.

gate tunneling, device physics

Leakage through gate dielectric.

gate-all-around (gaa) fet,gate-all-around,gaa,gaa fet,gaafet,gate all around,technology

Next-gen transistor with gate surrounding channel.

gate-first process, process integration

Gate-first processes deposit gate stack before source-drain formation simplifying but limiting thermal budget.

gate-first process,process

Form metal gate before source/drain.

gate-first vs gate-last, process integration

Impact on overall process flow.

gate-last (replacement gate),gate-last,replacement gate,process

Form gate after high-temperature processing.

gate-last process, process integration

Gate-last processes form gate after source-drain using replacement gate enabling high-k metal gate compatibility.

gate, packaging

Entry point to cavity.

gated convolution, llm architecture

Gated convolutions use multiplicative gates controlling information flow.

gated diode,metrology

Test structure for junction characterization.

gated fusion, multimodal ai

Use gates to control information flow.

gated linear layers, neural architecture

Combine linear transform with gating.

gatedcnn, neural architecture

Convolutional networks with gating.

gather-excite, computer vision

Efficient feature aggregation.

gating in transformers, transformer

Use gates to control information flow.

gating network,model architecture

Decides which experts to activate for each input.

gating networks, neural architecture

Decide which paths to activate.

gauge equivariant networks, scientific ml

Respect gauge symmetries.

gaussian approximation potentials, gap, chemistry ai

Kernel-based force field.

gaussian covariance, 3d vision

Shape and orientation of Gaussians.

gaussian process regression, data analysis

Model process with uncertainty quantification.

gaussian splatting training, 3d vision

Optimize Gaussian parameters.

gaussian splatting, 3d vision

Represent scenes as 3D Gaussians.

gaussian splatting, multimodal ai

Gaussian splatting represents scenes as collections of 3D Gaussians for real-time differentiable rendering.

gc-san, gc-san, recommendation systems

Graph Contextualized Self-Attention combines graph neural networks with self-attention for modeling both local and global session context.

gce-gnn, gce-gnn, recommendation systems

Global Context Enhanced Graph Neural Networks combine local session and global preference modeling.

gcn spectral, gcn, graph neural networks

Spectral graph convolutional networks define convolutions through graph Laplacian eigendecomposition.

gcpn, gcpn, graph neural networks

Graph Convolutional Policy Network generates graphs through reinforcement learning with domain-specific rewards for molecular design.

gdas, gdas, neural architecture search

Gradient-based architecture search with differentiable sampling enables efficient single-path supernet training.

gdpr,ccpa,data protection

GDPR/CCPA require consent, data minimization, right to deletion. Design AI systems with privacy by default.

gdsii format, gdsii, design

Standard design data format.

ge2e loss, ge2e, audio & speech

Generalized End-to-End loss trains speaker embeddings by optimizing similarity within and between speakers.

gedi (generative discriminator),gedi,generative discriminator,text generation

Use class-conditional language models for control.

geglu activation, neural architecture

GELU-gated linear unit.

gelu, neural architecture

Smooth activation used in Transformers.

gelu,swiglu,activation

GELU and SwiGLU are activation functions in transformers. SwiGLU used in modern LLMs like Llama.

gem300,automation

SECS/GEM standard for 300mm fabs.

geman-mcclure loss, machine learning

Robust loss for outliers.

gemba walk, manufacturing operations

Gemba walks involve management observing processes directly where work happens.