AI 용어집
인공지능 완전 사전
Denoising Diffusion Models
Fundamental architecture where data is progressively noised and then restored by learning the reverse process of diffusion.
Latent Diffusion Models
Approach applying diffusion in a reduced-dimensional latent space to improve computational efficiency.
Stochastic Score-Based Diffusion Models
Method based on estimating the data density gradient to guide the generation process.
Conditional Diffusion Models
Systems that generate data controlled by specific conditions such as text, images, or other modalities.
Diffusion by Stochastic Transformation
A theoretical framework unifying diffusion and score-based models based on stochastic differential equations.
Multi-Scale Diffusion Models
Architecture operating simultaneously on different spatial resolutions to capture details at multiple levels.
Continuous Learning Diffusion
Systems capable of adapting diffusion models without forgetting previously learned knowledge.
Classifier-Guided Diffusion Models
Technique using an external classifier to guide the generation process toward desired attributes.
Accelerated Sampling Diffusion
Methods optimizing the number of denoising steps to reduce generation time while preserving quality.
Multimodal Diffusion Models
Systems that generate and manipulate multiple types of data (text, image, audio, video) simultaneously.
Diffusion for 3D Generation
Specialized application of diffusion models for creating three-dimensional content and meshes.
Diffusion Models for Molecular Synthesis
Specialized systems for the generation and optimization of molecular structures for drug discovery.
Variable Step Diffusion
Architecture dynamically adapting the number and distribution of diffusion steps based on data complexity.
Hybrid Diffusion Models
Combination of diffusion models with other generative architectures such as GANs or VAEs.
Diffusion for Data Augmentation
Application of diffusion models to create realistic variations of existing training data.