29,258 technical terms and definitions
data layout transformation and cache hit rate improvement, data layout transformation, layout transformation cache, transformation cache hit, cache hit rate, hit rate improvement
Data Layout Transformation and Cache Hit Rate Improvement Introduction & Motivation Data Layout Transformation and Cache Hit Rate Improvement addresses a central problem in memory…
ai safety
Data Leakage is the critical machine learning vulnerability where information from outside the training dataset improperly influences model development — causing artificially…
Data Leakage Detection and Prevention Introduction & Motivation Data Leakage Detection and Prevention addresses a central problem in (subject) assurance and (target) management…
simd data parallelism, mimd task parallelism, instruction level parallelism, gpu vs cpu parallelism
Data-Level vs. Task-Level Parallelism represents the fundamental architectural and software design dichotomy that defines how programs divide immense computational workloads…
data line coupling prediction, data line coupling, line coupling prediction, data line, line coupling, coupling prediction
Data Line Coupling Prediction with Machine Learning Introduction & Motivation Data Line Coupling Prediction with Machine Learning addresses a central problem in data line…
dbt-Core Automated Data Lineage and DAG Testing Introduction & Motivation dbt-Core Automated Data Lineage and DAG Testing addresses a central problem in (subject) assurance…
Data Lineage and Provenance Tracking Introduction & Motivation Data Lineage and Provenance Tracking addresses a central problem in (subject) assurance and (target) management: how…
data lineage and provenance, data lineage provenance, data lineage, lineage provenance
Data Lineage and Provenance Introduction & Motivation Data Lineage and Provenance addresses a central problem in ml-ops process control and lineage-tracking-completeness…
infrastructure
Data loading pipeline is the end-to-end workflow that fetches, decodes, transforms, and delivers batches to accelerators during training - its job is to keep GPUs continuously fed…
data mesh architecture prediction fab quality analytics, data mesh, architecture prediction fab, sla, data mesh architecture, mesh architecture prediction, prediction fab quality
Data Mesh Architecture Prediction for Fab Quality Analytics Introduction & Motivation Data Mesh Architecture Prediction for Fab Quality Analytics addresses a central problem…
data migration strategy, data migration, migration strategy
Data Migration Strategy Introduction & Motivation Data Migration Strategy addresses a central problem in programming process control and migration-downtime-zero-downtime…
training techniques
Data Minimization is governance principle that limits collection and processing to data strictly necessary for defined purposes - It is a core method in modern semiconductor AI…
domain, proportion
Data mix balances training data across domains like web text books code and papers with proportions affecting model capabilities. Optimal mixing is empirically determined through…
training
Data mixing strategies is methods for combining multiple datasets into a single training mixture with controlled weighting - Mixing policies balance domain coverage, quality…
pretraining data composition, data ratio, domain weighting, training data curation
Pretraining Data Mixture and Curation is the strategic selection and weighting of training data domains that critically determines the capabilities, biases, and performance…
data normalization, preprocessing standardization scaling, iqr l2 nlp passed, data normalization & preprocessing standardization scaling & encoding, standardization scaling
Data Normalization & Preprocessing: Standardization, Scaling & Encoding Introduction & Motivation Preprocessing: prepare raw data for training. Standardization: zero mean, unit…
data offloading decision algorithm, data offloading decision, offloading decision algorithm, data offloading, offloading decision, decision algorithm
Data Offloading Decision Algorithm Introduction & Motivation Data Offloading Decision Algorithm addresses a central problem in edge-computing process control…
training
Data ordering effects is performance differences caused by the sequence in which training samples are presented - Even with identical data and compute, ordering can influence…
model parallel, hybrid
Data parallelism trains the same model on different data batches across multiple GPUs while model parallelism splits the model itself across GPUs. Hybrid approaches combine both…
ddp pytorch, distributed data parallel, data parallel training, allreduce training
Distributed Data Parallel (DDP) Training is the foundational parallelism strategy where the same model is replicated across multiple GPUs and each replica processes different data…
distributed data parallelism, gradient synchronization, ddp pytorch, batch size scaling
Distributed Data Parallelism (DDP) is the most widely-used distributed training strategy that replicates the entire model on every GPU and partitions the training data across GPUs…
map reduce parallel, stencil computation, embarrassingly parallel, parallel pattern language
Data Parallel Patterns are the fundamental computational building blocks — map, reduce, scan, gather, scatter, stencil, and histogram — that express common parallel operations on…
parallel map reduce scan, parallel primitives, collective operations
Data Parallel Patterns are the fundamental computational building blocks — map, reduce, scan, gather, scatter, stencil, and histogram — that express common parallel operations on…
distributed data parallel ddp, gradient synchronization, data parallel scaling, batch size scaling
Data Parallelism in Distributed Training is the most widely used distributed deep learning strategy where the model is replicated across N GPUs, each processing 1/N of…
model training
Data parallelism replicates the model on each device and processes different data batches in parallel. How it works: Copy complete model to each GPU, each processes different…
distributed data parallel, ddp training
Data parallelism is the simplest and most common way to scale training across many GPUs: replicate the entire model on every device, give each replica a different slice of…
data parallelism batch distribution and gradient synchronization, data parallelism batch, parallelism batch distribution, batch distribution gradient, data parallelism
Data Parallelism Batch Distribution and Gradient Synchronization Introduction & Motivation Data Parallelism Batch Distribution and Gradient Synchronization addresses a central…
ddp pytorch, zero redundancy optimizer, gradient compression, allreduce data parallel
Data Parallelism and Gradient Synchronization is the foundational distributed training approach where identical model replicas process different data samples, aggregate gradients…
ml data pipeline, etl, elt, data orchestration, airflow, dagster, prefect, spark, flink, kafka
Data pipeline is an automated, observable flow that ingests, validates, transforms, stores and delivers data products to downstream consumers. ML systems depend on reproducible…
input pipeline, prefetching data, data loader, io bound training
ML Data Pipeline is the system that efficiently loads, preprocesses, and batches training data — a bottleneck that can reduce GPU utilization from 100% to < 30% if poorly…
training, malicious
Data Poisoning is the adversarial attack that corrupts machine learning models by injecting malicious examples into training data — exploiting the fundamental dependence of ML…
ai safety
Data poisoning injects malicious samples into training data to corrupt model behavior. Attack goals: Untargeted: Degrade overall model performance. Targeted: Make model misbehave…
interpretability
Data Poisoning is a training-data attack that injects malicious or mislabeled samples to corrupt model behavior - It can degrade generalization or implant targeted failures while…
data prefetching accuracy prediction, data prefetching accuracy, prefetching accuracy prediction, data prefetching, prefetching accuracy, accuracy prediction
Data Prefetching Accuracy Prediction with Machine Learning Introduction & Motivation Data Prefetching Accuracy Prediction with Machine Learning addresses a central problem in data…
infrastructure
Data preprocessing at scale is the high-throughput transformation of raw datasets into model-ready tensors across large distributed environments - it must be engineered as…
training data extraction attack, membership inference attack
Data privacy, in the context of AI, is the problem of keeping information about real individuals from leaking out of a system that touched their data — not just during collection…
Data Profiling and Quality Assessment Introduction & Motivation Data Profiling and Quality Assessment addresses a central problem in (subject) assurance and (target) management…
training
Data proportions is the explicit percentage share of each dataset component within the final training corpus - Proportion settings control how often each data type contributes…
data lineage, Data Provenance Initiative, datasheets for datasets, Gebru 2018, dataset licensing, data cards, W3C PROV
Data provenance is the documented history of a dataset: where its contents came from, who collected or created them, under what terms, and what transformations, filters and merges…
validation, testing
Data Quality Data quality checks validate training data through schema validation distribution monitoring and anomaly detection because bad data produces bad models. Schema…
Composite Data Quality Index DQ-Score Formulation Introduction & Motivation Composite Data Quality Index DQ-Score Formulation addresses a central problem in (subject) assurance…
Multi-Dimensional Data Quality Assessment Introduction & Motivation Multi-Dimensional Data Quality Assessment addresses a central problem in (subject) assurance and (target)…
data quality validation, data quality, quality validation
Data Quality Validation Introduction & Motivation Data Quality Validation addresses a central problem in ml-ops process control and validation-latency-per-record prediction: how…
Data Reconciliation and Matching Introduction & Motivation Data Reconciliation and Matching addresses a central problem in (subject) assurance and (target) management: how…
data reordering and prefetch generation in compiler, data reordering prefetch, reordering prefetch generation, prefetch generation compiler, data reordering, reordering prefetch
Data Reordering and Prefetch Generation in Compiler Introduction & Motivation Data Reordering and Prefetch Generation in Compiler addresses a central problem in ml-compiler…
training
Data replay is reintroduction of selected past data during later training phases to preserve learned capabilities - Replay buffers protect important knowledge when models continue…
bits bytes binary hex, ascii unicode utf8 encoding, integer overflow two complement, ieee 754 floating point
A computer never actually stores a number, a letter, or a picture — it stores voltage levels that are either high or low, and every single thing a program manipulates is…
bits, bytes, binary, hexadecimal, text encoding, ASCII, Unicode, UTF-8, integers, integer overflow, floating-point, IEEE 754
A bit is the smallest unit of data a computer can hold: a single 0 or 1, physically stored as one of two voltage levels inside the RAM and CPU registers the first topic in this…
training techniques
Data Retention is policy framework that defines how long data is stored before deletion or archival - It is a core method in modern semiconductor AI serving and trustworthy-ML…
data scientist, Drew Conway data science Venn diagram, John Tukey future of data analysis, DJ Patil Jeff Hammerbacher, data science vs machine learning, data science workflow
Data science names the practice of extracting usable knowledge and decisions from data by combining three things that, on their own, are each insufficient: statistics to reason…