AI Factory Glossary

excursion response, production

Actions taken when excursion detected.

excursion,production

Process deviation or event that may affect quality.

executable semantic parsing,nlp

Parse to executable code or queries.

execution feedback,code ai

Run generated code get errors and iterate to fix.

execution trace, ai agents

Execution traces document complete agent behavior sequences for analysis.

executive order,biden,safety

US Executive Order on AI establishes safety standards. Reporting requirements for large models.

executive summary generation,content creation

Summarize documents for executives.

exemplar learning, self-supervised learning

Distinguish transformed versions of same image.

exemplar learning, self-supervised learning

Distinguish augmented versions.

exemplar selection,continual learning

Choose which examples to remember.

exfoliation, substrate

Splitting of implanted layer.

exhaust scrubber,facility

System to neutralize toxic gases from tools before venting.

exhaust system, manufacturing operations

Exhaust systems remove process byproducts and maintain tool vacuum.

exl2,exllama,efficient

EXL2 is quantization format for ExLlama. Dynamic bit allocation. Very fast GPU inference.

exllama,fast,gpu

ExLlama is fast LLM inference for GPU. Custom CUDA kernels. Efficient.

expanded uncertainty, metrology

Uncertainty with confidence interval.

expanding process window, process

Widen operating range.

expanding window, time series models

Expanding window forecasting retrains on all historical data including newest observations.

expectation over transformation, eot, ai safety

Evaluate robustness over transformations.

expected calibration error (ece),expected calibration error,ece,evaluation

Metric for calibration quality.

expediting, supply chain & logistics

Expediting accelerates delivery of critical materials through special handling or premium shipping.

experience curve, business

Broader learning effects.

experience replay for llms, continual learning

Replay old examples.

experience replay,continual learning

Randomly sample old examples during training.

experiment configuration management, mlops

Version and manage experiment configs.

experiment tracking,wandb,mlflow

Track experiments with W&B or MLflow: hyperparameters, metrics, artifacts. Reproduce and compare runs.

experiment,iterate,feedback loop

We can set up build-measure-learn loops: define hypotheses, metrics, experiments, and how to interpret results.

expert annotation,data

Use domain experts for high-quality labels.

expert capacity factor, moe

Maximum tokens per expert in MoE.

expert capacity, llm architecture

Expert capacity limits tokens assigned to each expert preventing overload.

expert capacity,moe

Maximum tokens each expert can process.

expert choice routing,moe

Experts choose which tokens to process rather than tokens choosing experts.

expert dropout, moe

Randomly disable experts during training.

expert load balancing, moe

Ensure even expert utilization.

expert parallel,moe,switch

Expert parallelism places MoE experts on different GPUs. Each token routed to subset of experts.

expert parallelism implementation, moe

Distribute experts across GPUs.

expert parallelism, llm architecture

Expert parallelism distributes experts across devices for scalable MoE training.

expert parallelism,distributed training

Distribute MoE experts across different GPUs.

expert redundancy, moe

Multiple experts learning similar features.

expert routing, llm architecture

Expert routing assigns tokens to appropriate specialized sub-networks.

expert routing,model architecture

Mechanism to select which experts process each input.

expert specialization, moe

Experts learning distinct features.

explainable ai for fab, data analysis

Interpret ML model decisions.

explainable recommendation,recommender systems

Provide reasons for recommendations.

explanation generation, recommendation systems

Explanation generation in recommendations provides users with interpretable reasons for why items were suggested.

explicit reasoning steps,reasoning

Show each reasoning step transparently.

exploration vs exploitation,reinforcement learning

Balance trying new actions vs using best known.

exploration-exploitation, recommendation systems

Exploration-exploitation balance in recommendations trades off exploiting known preferences versus discovering new interests.

exponential backoff, llm optimization

Exponential backoff increases delay between retry attempts preventing thundering herd.

exponential distribution, reliability

Constant failure rate model.