Expanding Window
time series models
Expanding Window is evaluation and training scheme where the historical window grows as time progresses. - It preserves all past data so long-run information remains available…
1,184 technical terms and definitions
time series models
Expanding Window is evaluation and training scheme where the historical window grows as time progresses. - It preserves all past data so long-run information remains available…
expectation maximization algorithm latent variables iterative optimization, expectation maximization, algorithm latent variables, 2d elbo em gmm hmm passed
Expectation-Maximization Algorithm: Latent Variables & Iterative Optimization Introduction & Motivation EM algorithm optimizes likelihood of models with latent variables…
expectation maximization and variants, expectation maximization variants, expectation maximization, maximization variants, bic em mnist, expectation-maximization variants
Expectation-Maximization and Variants Introduction & Motivation EM algorithm maximizes likelihood for models with latent variables. Fundamental for clustering, density estimation…
expectation maximization mixture model prediction, expectation maximization mixture, maximization mixture model, mixture model prediction, expectation maximization
Expectation Maximization Mixture Model Prediction with Machine Learning Introduction & Motivation Expectation Maximization Mixture Model Prediction with Machine Learning addresses…
eot, ai safety
EOT (Expectation Over Transformation) is a technique for attacking models that use stochastic defenses (randomized preprocessing, random dropout, random resizing) — computing…
expectation propagation distributed inference prediction, expectation propagation distributed, propagation distributed inference, distributed inference prediction
Expectation Propagation Distributed Inference Prediction with Machine Learning Introduction & Motivation Expectation Propagation Distributed Inference Prediction with Machine…
expected calibration error, ece, evaluation
Expected Calibration Error (ECE) is the primary metric for evaluating the calibration quality of a probabilistic classifier, measuring the average absolute difference between…
Expected Calibration Error ECE and Reliability Diagrams Introduction & Motivation Expected Calibration Error ECE and Reliability Diagrams addresses a central problem…
Expected Error Reduction via 0-1 Loss Minimization Introduction & Motivation Expected Error Reduction via 0-1 Loss Minimization addresses a central problem in {subject} learning…
Expected Error Reduction Query Strategy Introduction & Motivation Expected Error Reduction Query Strategy addresses a central problem in {subject} learning and efficient {target}…
Expected Model Output Change EMOC Query Strategy Introduction & Motivation Expected Model Output Change EMOC Query Strategy addresses a central problem in {subject} learning…
Expected Model Change Query Strategy Introduction & Motivation Expected Model Change Query Strategy addresses a central problem in {subject} learning and efficient {target}…
Expected Reciprocal Rank ERR Cascade Evaluation Introduction & Motivation Expected Reciprocal Rank ERR Cascade Evaluation addresses a central problem in recsys systems…
supply chain & logistics
Expediting is accelerated coordination actions used to recover delayed supply, production, or shipment commitments - It mitigates imminent service failure when normal lead-time…
business
Experience curve is the broader economic relationship where total cost declines with cumulative output due to scale and learning - Cost reductions come from process learning…
hindsight experience replay, reinforcement learning advanced
Hindsight Experience is goal-conditioned replay that relabels failed trajectories as successes for alternate achieved goals. - It extracts learning signal from unsuccessful…
continual learning
Experience replay is a technique from reinforcement learning — adopted for continual learning — where the model randomly samples and replays stored examples from previous…
continual learning, catastrophic forgetting, llm training, buffer replay, lifelong learning, ai
Experience replay is a continual-learning technique that reuses buffered past samples during training on new data - Replay batches interleave old and new examples so optimization…
Experience Replay Buffer Management Introduction & Motivation Experience Replay Buffer Management addresses a central problem in reinforcement policy learning…
iterate, feedback loop
Experimentation and Iteration The Build-Measure-Learn Loop For AI Applications Types of Experiments Prompt Experiments Test different system prompts Compare few-shot examples Try…
mlops
Experiment configuration management is the discipline of defining, versioning, validating, and governing all settings that determine experiment behavior - it prevents…
ml experiment tracking, weights and biases, wandb, mlflow, neptune, comet, tensorboard, clearml
Experiment tracking records the configuration, metrics, artifacts and provenance of model-development runs so results can be compared and reproduced. Without tracking, teams…
experiment tracking metadata, experiment tracking, tracking metadata
Experiment Tracking Metadata Introduction & Motivation Experiment Tracking Metadata addresses a central problem in ml-ops process control and tracking-overhead-performance…
factorial, randomization, block design, orthogonal array
Experimental design is the disciplined science of planning experiments so that the data they produce answer the engineer's question with maximum efficiency and minimum bias…
data
Expert annotation is the process of having domain specialists — such as doctors, lawyers, linguists, or engineers — create labeled training and evaluation data for machine…
moe
Expert Capacity is the maximum number of tokens that can be routed to any single expert within a Mixture-of-Experts (MoE) layer during a single forward pass, defined as…
architecture
Expert Capacity is maximum token budget assigned to each expert within a sparse mixture layer - It is a core method in modern semiconductor AI serving and inference-optimization…
moe
Expert Capacity Factor is the hyperparameter in Mixture of Experts (MoE) models that controls the maximum number of tokens each expert can process per batch — calculated as (total…
moe
Expert Choice Routing is the MoE routing paradigm that inverts the traditional token-selects-expert direction — instead, each expert independently selects the top-k tokens it…
moe
Expert dropout is the regularization technique that temporarily disables a subset of experts during training to reduce over-reliance on dominant experts - it encourages more…
moe
Expert load balancing is the process of distributing routed tokens across experts so no small subset becomes overloaded while others idle - it is essential for achieving both…
moe, switch
Expert parallelism distributes Mixture-of-Experts (MoE) model experts across different GPUs, enabling sparse activation with massive total parameter counts. Architecture: router…
distributed training
Expert parallelism is a distributed computing strategy specifically designed for Mixture of Experts (MoE) models, where different expert sub-networks are placed on different GPUs.…
architecture
Expert Parallelism is distributed execution strategy that shards experts across multiple devices for scalable sparse training - It is a core method in modern semiconductor AI…
moe
Expert parallelism implementation is the distributed execution strategy that shards experts across devices while sharing router work across replicas - it allows sparse models…
mixture experts parallelism, moe distributed training, expert placement strategies, load balancing experts
Mixture of Experts (MoE) is the sparse-activation architecture that scales a neural network to trillions of parameters while keeping per-token compute fixed — each input activates…
mixture of experts distributed, moe training parallelism, expert model parallel, switch transformer training
Mixture of Experts (MoE) is the sparse-activation architecture that scales a neural network to trillions of parameters while keeping per-token compute fixed — each input activates…
moe
Expert redundancy is the undesired condition where multiple MoE experts learn highly overlapping functions, reducing effective sparse capacity - it limits quality gains and wastes…
model architecture
Expert routing determines which experts process each token in Mixture of Experts architectures. Router network: Small network (often single linear layer) that takes token…
architecture
Expert Routing is process of assigning each token to one or more specialized experts in sparse architectures - It is a core method in modern semiconductor AI serving…
moe
Expert specialization is the emergent behavior where different MoE experts learn distinct token or task sub-distributions over training - it is the main mechanism by which sparse…
Large Language Model Training Modern LLM training follows a systematic approach from data to deployment: Training Pipeline Overview Large Language Model training is a multi-stage…
explainability interpretability shap lime attention saliency, explainability interpretability, shap lime attention, 10k ai bert cam clip compas, lime attention saliency
Explainability and Interpretability: SHAP, LIME, Attention, and Saliency 1. Introduction & Motivation As machine learning models become more powerful, they often become less…
Explainability Evaluation Metrics Introduction & Motivation Explainability Evaluation Metrics addresses a central problem in interpretability and explanation of xai predictions…
Explainability Evaluation Metrics (Extended Study) Introduction & Motivation Explainability Evaluation Metrics (Extended Study) addresses a central problem in interpretability…
Explainability Evaluation Metrics (Extended Study) Introduction & Motivation Explainability Evaluation Metrics (Extended Study) addresses a central problem in interpretability…
Explainability for Graph Neural Networks Introduction & Motivation Explainability for Graph Neural Networks addresses a central problem in interpretability and explanation of xai…
Explainability for Graph Neural Networks (Extended Study) Introduction & Motivation Explainability for Graph Neural Networks (Extended Study) addresses a central problem…
Explainability for Graph Neural Networks (Extended Study) Introduction & Motivation Explainability for Graph Neural Networks (Extended Study) addresses a central problem…
Explainability for Transformers Introduction & Motivation Explainability for Transformers addresses a central problem in interpretability and explanation of xai predictions: how…