EXPLORE / 06
Terms index
Specialized AI and software-development terms with concise explanations
RECORDS
Entity records
3372 records · 44/85
- 1721
Data Bias
Systematic errors or imbalances in training data that cause unfair model outputs
Website pending - 1722
Code Abstraction
A programming technique that hides complex implementation details behind simpler interfaces
Website pending - 1723
DRY Principle
A software development principle aiming to reduce repetition of software patterns
Website pending - 1724
CO2 Equivalent
A metric converting various greenhouse gases into their carbon dioxide equivalent impact
Website pending - 1725
AI ASIC
Integrated circuits custom-designed for AI workloads, such as Gaudi.
Website pending - 1726
Carbon Footprint
Total greenhouse gas emissions generated during computation or training processes
Website pending - 1727
BERT
Bidirectional Encoder Representations from Transformers for NLP tasks
Website pending - 1728
Model Reliability
The ability of a model to consistently perform tasks correctly under specific conditions.
Website pending - 1729
Model Trustworthiness
User confidence in the accuracy, fairness, and robustness of model outputs.
Website pending - 1730
Data Representativeness
The degree to which training data distribution mirrors the real-world application environment.
Website pending - 1731
Fully Sharded Data Parallel (FSDP)
A technique that shards model parameters, gradients, and optimizer states across GPUs to save memory
Website pending - 1732
Table Question Answering
NLP task extracting answers from structured tables
Website pending - 1733
Q-Learning
A model-free reinforcement learning method that finds the optimal policy by updating state-action value functions
Website pending - 1734
Deep Q-Learning (DQN)
A reinforcement learning algorithm combining deep learning with Q-learning to handle high-dimensional state spaces
Website pending - 1735
IPU
A processor architecture designed by Graphcore specifically for machine learning workloads.
Website pending - 1736
Denoising
The process of recovering original signals or images from noisy data.
Website pending - 1737
Dynamic Training
Model training process that adjusts strategies based on real-time feedback
Website pending - 1738
Adversarial Data
Inputs designed to mislead models into producing incorrect outputs
Website pending - 1739
Hardware Acceleration
Using specialized hardware to improve the execution efficiency of computational tasks.
Website pending - 1740
Model Serialization
Saving in-memory model structures to files for later loading and execution
Website pending - 1741
OOM
An error occurring when a program attempts to allocate more memory than is available.
Website pending - 1742
ZeRO
A distributed training technique that reduces memory footprint by sharding optimizer states, gradients, and parameters.
Website pending - 1743
Cosine Similarity
Measuring the angle between two vectors to assess semantic proximity
Website pending - 1744
TF Serving
An open-source system for serving machine learning models in production environments.
Website pending - 1745
A2C
A synchronous reinforcement learning algorithm combining policy gradient and value function estimation
Website pending - 1746
Linear Complexity
Complexity characteristic where runtime or memory grows linearly with input size.
Website pending - 1747
Nyström Method
An algorithm for low-rank approximation of kernel matrices or large symmetric matrices.
Website pending - 1748
Matrix Multiplication
Linear algebra operation that is the core computational task in neural network forward propagation
Website pending - 1749
8-bit Quantization
Compressing model weights or activations from high-precision floats to 8-bit integers to save memory
Website pending - 1750
Vision Transformer
An architecture that processes images by splitting them into patches and using a Transformer encoder
Website pending - 1751
Managed Service
Cloud-provided automated infrastructure and operations
Website pending - 1752
Open Source License
Legal agreements defining rights to use, modify, and distribute software or models
Website pending - 1753
Decision Transformer
Architecture modeling RL as sequence prediction
Website pending - 1754
Tensor Parallelism
Splitting single tensor operations across multiple devices
Website pending - 1755
AutoML
Techniques that automate tasks in machine learning workflows such as feature engineering and model selection.
Website pending - 1756
DOI
A persistent string system used to uniquely identify digital objects
Website pending - 1757
Evaluation Metrics
Statistical standards quantifying model performance or behavioral characteristics
Website pending - 1758
Trainer API
A high-level interface provided by Hugging Face that encapsulates the training loop.
Website pending - 1759
Decoding Strategy
Rules or algorithms for selecting the next token from model output distributions
Website pending - 1760
Contrastive Search
A decoding method that reduces repetition by balancing probability and diversity
Website pending