AI 语料库
专业术语解析 · 构建 AI 知识体系
#
Large Language Model
A deep learning language model trained on large-scale corpora, capable of text understanding, generation, and reasoning, with parameter scales typically ranging from billions to hundreds of billions.
Attention Mechanism
A technique that enables a model to dynamically attend to different parts of an input sequence by computing correlations between queries and keys/values to perform information weighting.
Fine-tune
A technique that fine-tunes a pre-trained model using domain-specific data to adapt it for downstream tasks.
Object Detection
Object detection is a core task in computer vision that not only identifies objects present in an image but also precisely locates them, typically using bounding boxes. YOLO and Faster R-CNN are the leading object detection algorithms.
Reinforcement Learning
Reinforcement Learning (RL) is one of the three core machine learning paradigms. It learns optimal policies through agent-environment interaction, where agents continuously adjust their strategies based on reward feedback to maximize cumulative rewards.
Gradient Descent
Gradient descent is the most fundamental optimization algorithm in machine learning. It iteratively updates parameters along the negative gradient of the loss function to minimize the loss value. It is the core method for training neural networks.
Image Segmentation
Image segmentation divides an image into multiple regions with semantic meaning. It includes semantic segmentation (pixel-level classification) and instance segmentation (distinguishing individual objects of the same class), and is widely used in medical imaging analysis and autonomous driving.
Word Embedding
Word embeddings map words to low-dimensional continuous vector spaces, placing semantically similar words closer together. Word2Vec and GloVe are classic embedding methods that laid the foundation for subsequent language models.
Named Entity Recognition
Named Entity Recognition (NER) is a foundational task in natural language processing that identifies meaningful entities—such as people, locations, organizations, and time expressions—from unstructured text. It is a key technology for information extraction and knowledge graph construction.
Edge Computing
Edge computing is a computing model that processes and analyzes data at the network edge, close to its source. In AI scenarios, it deploys inference capabilities to end devices to reduce latency and protect privacy, serving as a core enabling technology for smart hardware.