AI 용어집
인공지능 완전 사전
GANs (Generative Adversarial Networks)
Adversarial neural networks that generate realistic synthetic data by having a generator and a discriminator compete.
VAEs (Variational Autoencoders)
Variational autoencoders that learn a latent distribution to generate new samples by sampling in this space.
Diffusion Models
Progressive generative approach that adds and then removes noise from data to create high-quality samples.
Generative Transformers
Attention-based architecture for generating sequential data such as text, code, or time series.
Geometric Data Augmentation
Spatial transformations applied to images (rotation, translation, scaling) to increase the diversity of training data.
Mixup and Blending Techniques
Linear combination of existing samples to create new training instances with interpolated labels.
Temporal Data Synthesis
Generation of time series and sequential data to compensate for the lack of historical data.
Graph Generation
Synthetic creation of graph and network structures for training models on relational data.
Advanced Audio Synthesis
Generation of realistic audio data including speech, music, and sound effects to enrich datasets.
3D and Volumetric Generation
Creation of synthetic three-dimensional data such as point clouds, meshes, and medical volumes.
Synthetic Domain Adaptation
Targeted data generation to adapt models to new domains or conditions not represented in the original data.
Multi-Modal Synthesis
Coordinated generation of data from multiple modalities (image-text, audio-video) to create coherent sets.