27,049 technical terms and definitions
dsm systems design, virtual shared address space, memory coherence distributed, scalable shared memory
Distributed Shared Memory Architecture — DSM systems provide a shared memory abstraction over physically distributed memory nodes, enabling transparent data access across…
memory consistency model, coherence protocol, dsm system
Distributed Shared Memory (DSM) and Consistency Models define how memory operations across multiple processors are ordered and made visible to other processors, establishing…
Distributed Consensus, Paxos, Raft, and Byzantine Replication Distributed consensus allows independent machines to behave like one reliable state machine despite delay…
distributed computing, cap theorem, consensus protocol, fault tolerance, distributed training
Distributed systems is the engineering discipline of building reliable, scalable computing services across multiple networked machines that coordinate to appear as a single…
mlops
Distributed tracing is an observability technique that tracks a single request as it flows through multiple services in a distributed system, recording timing, metadata…
distributed tracing correlation, distributed tracing, tracing correlation
Distributed Tracing Correlation Introduction & Motivation Distributed Tracing Correlation addresses a central problem in ml-ops process control…
distributed tracing latency, distributed tracing, tracing latency
Distributed Tracing Latency Introduction & Motivation Distributed Tracing Latency addresses a central problem in cloud process control and trace-sampling-rate-accuracy prediction…
ddp, fsdp
Distributed Training Training Paradigms Data Parallel (DDP) Each GPU has full model copy, processes different data: Model Parallel Split model across GPUs: Tensor Parallel: Split…
data parallel training, ddp, fsdp, tensor parallel, pipeline parallel, expert parallel, nccl
Distributed training coordinates optimization across multiple accelerators and often multiple nodes. Large language, vision, recommendation, scientific, and multimodal models…
data parallel training pytorch, ddp distributed data parallel, gradient synchronization training, data parallel scaling efficiency
Data Parallel Distributed Training is the most widely used strategy for scaling deep learning training across multiple GPUs or nodes by replicating the entire model on each…
horovod distributed, pytorch distributed, deepspeed training, distributed ml framework
Distributed Training Frameworks are the software systems that coordinate the training of large machine learning models across multiple GPUs and multiple machines — handling data…
hierarchical all-reduce algorithm, multi-level allreduce
Hierarchical all-reduce is the two-level collective strategy that reduces gradients within nodes first, then across nodes - it exploits faster intra-node links and minimizes…
distributed training scalability and communication, distributed training scalability, training scalability communication, distributed training, training scalability
Distributed Training Scalability and Communication Introduction & Motivation Distributed Training Scalability and Communication addresses a central problem in ml-training process…
weak strong scaling analysis, communication overhead scaling, parallel efficiency metrics, scalability bottlenecks
Distributed Training Scaling Efficiency is the measure of how effectively training performance improves with additional compute resources — quantified through strong scaling…
distributed training synchronization, distributed training, training synchronization
Distributed Training Synchronization Introduction & Motivation Distributed Training Synchronization addresses a central problem in systems process control…
distributed transaction commit, distributed transaction, transaction commit
Distributed Transaction Commit Introduction & Motivation Distributed Transaction Commit addresses a central problem in distributed-systems process control…
semi-supervised learning
Distribution Alignment is a technique in semi-supervised learning that adjusts pseudo-label distributions to match the true class distribution — preventing the model from being…
robustness
Distribution Shift is the discrepancy between the data distribution during training and the distribution encountered during deployment — when $P_{test}(X, Y) eq P_{train}(X, Y)$…
ai safety
Distribution Shift is the change between training-time data distribution and real-world deployment data over time or context - It is a core method in modern AI safety execution…
reinforcement learning advanced
Distributional Bellman is Bellman operators over full return distributions instead of only expected scalar value. - It models uncertainty and multimodal outcomes that…
Distributional Reinforcement Learning (C51) Introduction & Motivation Distributional Reinforcement Learning (C51) addresses a central problem in sequential decision making via…
Distributional Reinforcement Learning (C51) (Extended Study) Introduction & Motivation Distributional Reinforcement Learning (C51) (Extended Study) addresses a central problem…
Distributional Reinforcement Learning (C51) (Extended Study) Introduction & Motivation Distributional Reinforcement Learning (C51) (Extended Study) addresses a central problem…
Distributional Reinforcement Learning Introduction & Motivation Distributional Reinforcement Learning addresses a central problem in reinforcement policy learning…
reinforcement learning
Distributional RL is a reinforcement learning framework that learns the full distribution of returns (not just the expected value) — instead of estimating , distributional RL…
Distributional RL and Quantile Regression Introduction & Motivation Distributional RL and Quantile Regression addresses a central problem in reinforcement policy learning…
Distributional Robustness Optimization Introduction & Motivation Distributional Robustness Optimization addresses a central problem in uncertainty methods and prediction…
distributionally robust optimization (dro) prediction, distributionally robust optimization, robust optimization (dro), optimization (dro) prediction, robust optimization
Distributionally Robust Optimization (DRO) Prediction with Machine Learning Introduction & Motivation Distributionally Robust Optimization (DRO) Prediction with Machine Learning…
distributor, channel partners, resellers, partners, distribution
Yes, we work with select distributors and channel partners to extend our global reach and provide local support — partnering with authorized distributors including Arrow…
defects
Divacancy is the point defect complex formed by two adjacent vacancies sharing a common lattice region — more thermally stable than isolated vacancies, it introduces deep…
code ai
Divergent Change is a code smell where a single class is frequently modified for multiple different, unrelated reasons — making it the collision point for changes originating…
text generation
Diverse beam search is the beam-search variant that adds diversity penalties across beams to generate multiple distinct high-quality hypotheses - it addresses beam collapse into…
optimization
Diverse Beam Search is beam-search variant that adds diversity penalties to produce distinct candidate outputs - It is a core method in modern semiconductor AI serving…
Diversity-Driven Skill Discovery Introduction & Motivation Diversity-Driven Skill Discovery addresses a central problem in reinforcement policy learning and skill-diversity-metric…
recommender systems
Diversity in recommendations ensures variety in suggested items — balancing relevance with diversity to avoid filter bubbles, expose users to different types of content…
reinforcement learning advanced
Diversity Intrinsic is intrinsic-reward design that encourages agents to learn behaviorally distinct skills. - It promotes broad state-space coverage and avoids collapsing to one…
Diversity in Recommendation Systems Introduction & Motivation Diversity in Recommendation Systems addresses a central problem in recsys systems and {target} optimization: how…
recommendation systems
Diversity regularization is regularization techniques that encourage recommendation lists to contain varied items or attributes - Additional loss terms penalize redundancy…
prompting techniques
Diversity Sampling is a selection strategy that prioritizes varied examples to cover multiple patterns within limited context budget - It is a core method in modern LLM execution…
Diversity-Based Sampling in Active Learning Introduction & Motivation Diversity-Based Sampling in Active Learning addresses a central problem in {subject} learning and efficient…
divide-and-conquer analysis
Divide-and-Conquer Analysis Introduction & Motivation Divide-and-Conquer Analysis addresses a central problem in algorithms process control and recursion-depth-logarithmic…
video understanding
Divided Space-Time Attention is a computationally efficient factorization strategy for Video Vision Transformers that decomposes the prohibitively expensive Joint Space-Time…
divider chain phase error accumulation prediction, divider chain phase, chain phase error, phase error accumulation, error accumulation prediction, divider chain, chain phase
Divider Chain Phase Error Accumulation Prediction with Machine Learning Introduction & Motivation Divider Chain Phase Error Accumulation Prediction with Machine Learning addresses…
python, batteries
Django is the batteries-included Python web framework that provides ORM, admin interface, authentication, and security features out of the box — used in AI applications requiring…
delay, locked, loop, design, fine, tuning
DLL: Delay-Locked Loop Design and Fine-Grained Delay Tuning is feedback circuits controlling propagation delay to match reference — enabling precision clock distribution, timing…
dlp data loss prevention, dlp data loss, data loss prevention, dlp data, data loss, loss prevention
DLP Data Loss Prevention Introduction & Motivation DLP Data Loss Prevention addresses a central problem in cloud process control and detection-accuracy-false-positives prediction…
deep level transient spectroscopy, deep level spectroscopy, semiconductor trap spectroscopy, trap activation energy, defect characterization dlts
Deep-level transient spectroscopy identifies electrically active defects by watching a semiconductor junction return toward equilibrium after a controlled filling pulse.…
direct memory access, scatter gather dma, sg dma, cyclic dma, bus master, iommu, gpudirect dma
DMA controller moves blocks of data between memory and peripherals or memory regions without CPU copy loops. DMA is essential for high-throughput storage, networking, audio…
dma direct memory access, dma direct memory, direct memory access, dma direct, direct memory, memory access
DMA Direct Memory Access Introduction & Motivation DMA Direct Memory Access addresses a central problem in systems process control and dma-transfer-efficiency-bandwidth…
zero copy transfer, direct memory access, gpu dma, memory transfer engine
DMA Engines and Zero-Copy Transfers are the hardware components and programming techniques that transfer data between memory regions (CPU↔GPU, GPU↔NVMe, NIC↔GPU) without CPU…