infant defect,manufacturing
Defect caught in early testing.
9,967 technical terms and definitions
Defect caught in early testing.
Early failure period.
Infant mortality is elevated early failure rate from latent defects eliminated by burn-in.
Early failures due to defects.
Computational expense of generating outputs after training.
Inference: Running the trained model to generate outputs. Optimize with quantization, batching, KV-cache. Cost = tokens x compute per token.
Attention with infinite context.
High-bandwidth low-latency network for HPC.
InfiniBand provides low-latency RDMA networking. Essential for distributed training. 200-400 Gbps.
Infinite capacity scheduling assumes unlimited resources identifying bottlenecks and capacity needs.
Neural network behavior as width approaches infinity.
Influence functions estimate impact of training examples on test predictions.
Influence functions identify training examples most influential to recommendations enabling interpretability and debugging.
Measure training example impact on predictions.
Influence propagation models how preferences spread through social networks.
TSMC's fan-out wafer-level packaging.
Disentangle GAN latents using mutual information.
Create visual information graphics.
Contrastive loss maximizing mutual information.
Maximize information about environment.
Informer uses ProbSparse self-attention to reduce computational complexity enabling efficient long sequence time series forecasting.
Use IR to see through wafer.
Ellipsometry in mid-IR range.
IR imaging for thermal analysis.
Infrared sensors measure temperature non-contact through thermal radiation.
Inhibitory point processes model negative influence where events decrease likelihood of subsequent events.
Inhomogeneous Poisson processes have time-varying intensity rates modeling non-stationary event occurrence patterns.
Inject liquid compound into mold.
Mark with ink.
Inking marks known-good or known-bad die with colored dots during wafer sort enabling visual identification for subsequent processing.
Check wafers during processing.
Inline metrology correlates process measurements with final yield identifying contributing factors.
Continuous measurement during production.
Inline yield tracks functional or parametric yield after specific process steps enabling early detection of yield excursions.
Remove bias information from representations.
Fill masked regions as pretext task.
Inpainting fills masked regions coherently using diffusion models conditioned on surrounding context.
Specify region to modify.
Fill masked regions.
Fill in missing or masked regions of images.
Inpainting fills masked regions. Outpainting extends beyond borders. AI image editing.
Simple attribution method.
Input filters screen prompts for policy violations or attacks.
Input reduction iteratively removes tokens while maintaining predictions identifying necessary context.
Clean and validate user inputs before sending to LLM to prevent attacks.
Validate and sanitize user inputs. Filter profanity, PII, malicious patterns before sending to model.
Check and sanitize user inputs.
Vary compute per input.
Input-dependent depth networks use different layer counts per sample based on complexity.
Insertion delay is delay through clock distribution network from tree root.