← Back to AI Factory Chat

AI Factory Glossary

751 technical terms and definitions

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z All
Showing page 5 of 16 (751 entries)

membership inference attack,ai safety

Determine if specific data was in the training set.

membership inference attacks, privacy

Determine if example was in training.

membership inference, interpretability

Membership inference determines whether specific examples were in training set revealing privacy leaks.

membership inference,privacy

Determine if example was in training set.

membership inference,privacy,attack

Membership inference determines if sample was in training data. Privacy attack. Differential privacy defends.

membrane filtration, environmental & sustainability

Membrane filtration removes particles and microorganisms through size-exclusion using porous membranes.

memit, memit, model editing

Edit multiple facts simultaneously.

memorizing transformer,llm architecture

Retrieve from large external memory.

memory bandwidth, business & strategy

Memory bandwidth measures data transfer rate between processor and memory.

memory bandwidth,hardware

Rate of data transfer to/from GPU memory often bottleneck for inference.

memory bank, self-supervised learning

Store representations for contrastive learning.

memory bist, advanced test & probe

Memory BIST uses on-chip logic to apply test algorithms like March or checkerboard patterns for memory array fault detection.

memory bist, design & verification

Memory BIST generates test algorithms for embedded memories.

memory buffer,continual learning

Store representative examples from old tasks.

memory coalescing, optimization

Efficient memory access patterns.

memory coalescing,access pattern

Coalesced memory access: adjacent threads access adjacent memory. Maximizes bandwidth. Critical for GPU performance.

memory consolidation, ai agents

Memory consolidation transfers important information from working to long-term storage.

memory in language models, theory

How models store information.

memory networks,neural architecture

Networks with explicit external memory for facts.

memory pool, llm optimization

Memory pools preallocate buffers reducing allocation overhead during inference.

memory profile,leak,allocation

Memory profiling tracks allocations. Find leaks, reduce footprint. Python memory_profiler, tracemalloc.

memory profiling, optimization

Analyze memory usage patterns.

memory redundancy, yield enhancement

Memory redundancy incorporates spare rows and columns that can replace defective elements improving yield through post-manufacturing repair.

memory retrieval agent, ai agents

Memory retrieval selects relevant past information based on current context.

memory retrieval, dialogue

Access relevant past messages.

memory stacking,advanced packaging

Vertically stack memory dies for higher density.

memory summarization, dialogue

Compress conversation history.

memory systems,ai agent

Give agents long-term persistent memory across sessions using vector stores or databases.

memory transformer-xl,llm architecture

Cache previous segments for longer context.

memory update gnn, graph neural networks

Memory update mechanisms in temporal GNNs maintain node states updated by events and read for predictions.

memory wall,bandwidth bottleneck

Memory wall: compute outpaces memory bandwidth. AI models are memory-bound. HBM and on-chip memory help.

memory-augmented video models, video understanding

Use external memory for long videos.

memory-bound operations, model optimization

Memory-bound operations are limited by data transfer rather than computation affecting actual deployment speed.

memory-bound vs compute-bound operations, optimization

Bottleneck type.

memory-efficient attention patterns, optimization

Attention variants saving memory.

memory-efficient training techniques, optimization

Methods to reduce memory usage.

memory,conversation history,context

Agent memory stores conversation history, facts, summaries. Enables multi-turn conversations and long-term context.

memory,kvcache,attention cache

KV-cache stores past keys/values so the model does not recompute them every token. Dramatically speeds up long-context inference with transformers.

memory,long term,persist

LLM memory persists facts across sessions. Vector DB or structured storage. Personalization.

memristors, research

Resistive memory devices.

mems fabrication, mems, process

Microelectromechanical systems processing.

mems packaging, mems, packaging

Protect mechanical structures.

mems probe card, mems, advanced test & probe

MEMS probe cards use microfabricated structures for ultra-fine pitch testing with improved planarity contact force control and probe density.

mentor,advisor,career

Find mentors in AI field. Learn from their experience. Give back by mentoring others.

meol, meol, process integration

Middle-End-Of-Line encompasses processes between transistor formation and traditional metal interconnect layers.

mercury porosimetry, metrology

Use mercury intrusion to measure pores.

mercury probe, metrology

Contact-based electrical testing on wafer.

merge lot, manufacturing operations

Merge operations combine split lots rejoining wafers into single lot.

merging,model merge,soup

Model merging combines weights from multiple fine-tuned models. Can get benefits of each without retraining.

mes (manufacturing execution system),mes,manufacturing execution system,production

Software managing and tracking fab operations.