hyena
llm architecture
Hyena is a subquadratic attention replacement that combines long convolutions (computed via FFT) with element-wise data-dependent gating — achieving O(n log n) complexity instead…
794 technical terms and definitions
llm architecture
Hyena is a subquadratic attention replacement that combines long convolutions (computed via FFT) with element-wise data-dependent gating — achieving O(n log n) complexity instead…
architecture
Hyena Hierarchy is long-sequence architecture using implicit long convolutions and hierarchical filtering operators - It is a core method in modern semiconductor AI serving…
what is hyper-threading, hyper-threading definition
Hyper-Threading is Intel’s implementation of simultaneous multithreading. A supported physical core exposes more than one architectural execution context to software.…
hyperband bandit feedback and multi-armed bandit framework, hyperband bandit feedback, bandit feedback multi-armed, feedback multi-armed bandit, multi-armed bandit framework
Hyperband Bandit Feedback and Multi-Armed Bandit Framework Introduction & Motivation Hyperband Bandit Feedback and Multi-Armed Bandit Framework addresses a central problem…
Hyperband Bandit-Based Early Stopping Strategy Introduction & Motivation Hyperband Bandit-Based Early Stopping Strategy addresses a central problem in optimization algorithm…
neural architecture search
Hyperband NAS is resource-allocation strategy using successive halving to evaluate many architectures efficiently. - It starts broad with cheap budgets and progressively focuses…
Hyperband Multi-Fidelity Optimization Introduction & Motivation Hyperband Multi-Fidelity Optimization addresses a central problem in optimization algorithm tuning and improvement…
Hyperbolic Collaborative Filtering in Poincare Ball Introduction & Motivation Hyperbolic Collaborative Filtering in Poincare Ball addresses a central problem in recsys systems…
Hyperbolic Graph Neural Network Embeddings Introduction & Motivation Hyperbolic Graph Neural Network Embeddings addresses a central problem in graph network representation…
weight generation, meta network, hypernetwork neural, dynamic weight generation
Hypernetworks are the neural networks that generate the weights of another neural network — where a small "hypernetwork" takes some conditioning input (task description…
HyperNetworks for Direct Few-Shot Weight Generation Introduction & Motivation HyperNetworks for Direct Few-Shot Weight Generation addresses a central problem in fewshot…
neural architecture
Hypernetworks are neural networks that generate the weights of another neural network — a meta-architectural pattern where a smaller "hypernetwork" produces the parameters of…
generative models
Hypernetworks for diffusion is the auxiliary networks that generate or modulate weights in diffusion layers to alter style or concept behavior - they provide an alternative…
bayesian, tune
Hyperopt is a Python library for Bayesian hyperparameter optimization — intelligently searching the hyperparameter space using probabilistic models to find optimal configurations…
tuning, sweep
A hyperparameter is any training setting a human chooses before training starts, as opposed to a parameter — a weight the training process itself learns by gradient descent.…
tuning, sweep
Hyperparameter tuning systematically searches for optimal values of learning rate, batch size, regularization, and architecture choices, using grid search, random search, Bayesian…
bayesian optimization, hpo, learning rate search, hyperparameter tuning
Hyperparameter Optimization (HPO) is the systematic search for the best configuration of training settings (learning rate, batch size, architecture choices, regularization) that…
cifar ei nas passed pbt
Hyperparameter Optimization Introduction & Motivation Hyperparameter Optimization: automatically tune model hyperparameters. Bayesian optimization, grid search, random search.…
automl, neural architecture search, bayesian optimization, automated machine learning
Hyperparameter Optimization and AutoML — Automating the Design of Deep Learning Systems Hyperparameter optimization (HPO) and Automated Machine Learning (AutoML) systematically…
optuna hyperparameter tuning, population based training, hyperparameter search neural network, bayesian optimization hpo
Hyperparameter Optimization (Bayesian, Optuna, Population-Based Training) is the systematic process of selecting optimal training configurations—learning rates, batch sizes…
optuna hyperparameter tuning, ray tune distributed, bayesian optimization deep learning, hpo automated tuning
Hyperparameter Optimization (HPO) is the systematic process of selecting the best configuration of training hyperparameters — learning rate, batch size, architecture choices…
hyperparameter optimization, bayesian optimization hyperband, bo ei gp hpo lr mnist, hyperparameter optimization bayesian optimization & hyperband, optimization bayesian
Hyperparameter Optimization: Bayesian Optimization & Hyperband Introduction & Motivation Hyperparameter optimization: automated search for best hyperparameters. Grid search…
hyperparameter optimization, bayesian random grid, bert bo cifar cv ei gp, hyperparameter optimization bayesian random & grid search, hyperparameter optimization bayesian, random grid
Hyperparameter Optimization: Bayesian, Random & Grid Search Introduction & Motivation Hyperparameter optimization: find best model configuration. Grid search: exhaustive parameter…
hyperparameter optimization bayesian search and surrogate modeling, hyperparameter optimization bayesian, optimization bayesian search, bayesian search surrogate, bayesian search
Hyperparameter Optimization Bayesian Search and Surrogate Modeling Introduction & Motivation Hyperparameter Optimization Bayesian Search and Surrogate Modeling addresses a central…
bayesian hyperparameter tuning, neural architecture search automl, hyperband successive halving, optuna hpo
Hyperparameter Optimization (HPO) is the automated search for the optimal configuration of neural network training hyperparameters (learning rate, batch size, weight decay…
mlops
Hyperparameter tracking is the structured recording and analysis of tuning parameter choices and their performance outcomes - it enables data-driven optimization by revealing…
model training
Hyperparameter tuning searches for optimal training settings like learning rate, batch size, and architecture choices. What are hyperparameters: Settings not learned by training…
hyperparameter optimization, grid search, random search
Hyperparameter tuning is the process of selecting configuration values that control how a machine-learning model is trained and generalized, such as learning rate, regularization…
hyperparameter tuning search, hyperparameter tuning, tuning search
Hyperparameter Tuning Search Introduction & Motivation Hyperparameter Tuning Search addresses a central problem in systems process control and search-strategy-grid-random-bayesian…
hyperscale data center energy efficiency optimization, hyperscale data center, data center energy, center energy efficiency, energy efficiency optimization, hyperscale data
Hyperscale Data Center Energy Efficiency Optimization Introduction & Motivation Hyperscale Data Center Energy Efficiency Optimization addresses a central problem in data-center…
hyperscale data center efficiency hardware, hyperscale data center, data center efficiency, center efficiency hardware, hyperscale data, data center, center efficiency, efficiency hardware
Hyperscale Data Center Efficiency Hardware Introduction & Motivation Hyperscale Data Center Efficiency Hardware addresses a central problem in data-center process control and pue…
hyperscale data center energy efficiency, hyperscale data center, data center energy, center energy efficiency, hyperscale data, data center, center energy, energy efficiency
Hyperscale Data Center Energy Efficiency Introduction & Motivation Hyperscale Data Center Energy Efficiency addresses a central problem in data-center process control…
hyperscale data center energy efficiency, hyperscale data center, data center energy, center energy efficiency, hyperscale data, data center, center energy, energy efficiency
Hyperscale Data Center Energy Efficiency Introduction & Motivation Hyperscale Data Center Energy Efficiency addresses a central problem in data-center process control and pue…
hyperspectral cathodoluminescence, cathodoluminescence spectrum imaging, cl hyperspectral mapping, cl data cube, hyperspectral cl semiconductor
A conventional cathodoluminescence image assigns one brightness value to each electron-beam position. Hyperspectral cathodoluminescence keeps the spectrum instead. Every raster…
hyperspectral imaging defect classification accuracy prediction, hyperspectral imaging defect, imaging defect classification, hyperspectral imaging
Hyperspectral Imaging Defect Classification Accuracy Prediction with Machine Learning Introduction & Motivation Hyperspectral Imaging Defect Classification Accuracy Prediction…
Hyperspectral Image Super-Resolution Introduction & Motivation Hyperspectral Image Super-Resolution addresses a central problem in vision processing…
quality & reliability
Hypothesis Test is a formal decision framework for evaluating evidence against a baseline process assumption - It is a core method in modern semiconductor statistical analysis…
Ronald Fisher Neyman Pearson, Type I Type II error, significance level, replication crisis p-hacking
Hypothesis testing is a formal statistical procedure for deciding whether an observed pattern in data — a difference between two groups, a relationship between two variables — is…
rag
Hypothetical Document Embeddings (HyDE) improves retrieval-augmented generation by using an LLM to generate a hypothetical answer to a query then embedding that hypothetical…
ai safety
Hypothetical scenarios is the prompt framing technique that presents harmful or restricted requests as theoretical questions to reduce refusal likelihood - it tests whether safety…
hafnium dioxide HfO2, equivalent oxide thickness EOT, gate tunneling leakage, metal gate TiN TaN, HKMG stack, Fermi level pinning, poly depletion, 45nm node Intel, gate capacitance scaling
High-k dielectric refers to a gate insulator material with a dielectric constant significantly higher than silicon dioxide's roughly 3.9, and its introduction solved a genuinely…
HBM stack, high bandwidth memory packaging, HBM3E packaging, HBM4 packaging, TSV DRAM stack, hbm
High-bandwidth memory (HBM) is DRAM stacked vertically and connected to its host through an interface thousands of bits wide, running beside the processor on an interposer rather…
ha system, active active, active passive, multi region, failover, availability nines, error budget
High availability designs a service to remain usable through routine failures, maintenance and demand changes within a stated availability objective. AI serving needs redundant…
A high-k metal gate is a transistor structure that replaces the traditional silicon dioxide gate insulator and polysilicon gate electrode with a high-dielectric-constant…