Glosarium AI
Kamus lengkap Kecerdasan Buatan
Graph Generative Models
Neural network architecture designed to learn the underlying distribution of graphs and generate new graph structures with properties similar to the training data.
GraphVAE
Variational Autoencoder adapted for graphs using a continuous latent space to learn a probabilistic distribution on graph structures and enable sampling of new graphs.
GraphGAN
General Adversarial Network applied to graphs where a generator creates graph structures and a discriminator evaluates their authenticity compared to real graphs.
GraphRNN
Sequential generative model using recurrent neural networks to generate graphs node by node by modeling the distribution of edge addition sequences.
Molecular Graph Generation
Specialized application of graph generative models for creating new valid molecular structures with desired chemical properties.
Graph Autoregressive Models
Generative approach decomposing the joint probability of a graph into a product of conditional probabilities to sequentially generate nodes and edges.
Graph Flow-based Models
Generative models using bijective transformations to map between the graph space and a simple latent space, enabling exact generation and density estimation.
Graph Diffusion Models
Generative models that progressively apply noise to training graphs and then learn to reverse this process to generate new graph structures.
Graph Normalizing Flows
Series of invertible transformations on graphs enabling exact distribution modeling and efficient sampling of new graphs.
Graph-based VAE
Variational Autoencoder specifically designed to handle the permutation-invariant nature of graphs and their complex topological structure.
Graph Convolutional Networks for Generation
Use of GNNs as encoders or decoders in generative models to capture structural dependencies during graph generation.
Graph Attention Networks for Generation
Attention mechanism adapted for graphs to selectively weight the influence of different parts of the graph during the generative process.
Graph Transformers for Generation
Modified transformer architecture to process graph structures by incorporating structural biases and graph-specific attention mechanisms.
Graph Reinforcement Learning for Generation
Approach treating graph generation as a sequential decision-making process where an agent learns to build optimal graphs through trials and rewards.
Graph Energy-based Models
Generative models defining an energy function on graphs where low-energy graphs are more probable, using MCMC-type sampling.
Graph Implicit Models
Generative models implicitly defining the distribution of graphs without explicit form, using techniques like GANs or score-based models.
Graph Neural ODEs
Ordinary differential equations applied to graphs to model continuous dynamics in latent space and generate graphs through numerical resolution.
Graph Variational Inference
Approximate inference technique adapted for graphs to estimate posterior distributions in probabilistic generative graph models.
Graph Latent Space Models
Representation of graphs in a low-dimensional latent space where distances and geometric relationships encode the topological structure of the original graph.
Graph Sequential Generation
Generative paradigm that builds graphs step by step, typically by sequentially adding nodes and their connections according to a learned policy.