Variational Autoencoder
A generative deep learning model that introduces latent variables following a probability distribution between the encoder and decoder, enabling probabilistic modeling of data and diverse sample generation. It is widely applied to text, image, and other generative tasks.
Timeline (last 90 days)
Generative molecular models (such as variational autoencoders, diffusion models and reinforcement learning strategies) can directly design new molecules that satisfy specific properties, and then molecular dynamics simulations predict their binding stability with the target protein.
A variational autoencoder (VAE) contains an encoder and a decoder, samples latent variables through the reparameterization trick, and its loss function consists of a reconstruction error and a KL divergence regularization term.
Amortized inference originates from the design paradigm of variational autoencoders (VAEs), using a shared inference network to reduce the parameter count from O(n) to the network's fixed parameter count, while enabling one-step inference generalization to new observation points.
Molecular generation and optimization can leverage graph neural networks (GNN), variational autoencoders (VAE), or diffusion models to generate entirely new molecular structures with specific properties.
Variational response generators introduce the idea of variational autoencoders by inserting a latent variable between the encoder and decoder; the latent variable follows a probability distribution and is sampled during generation to introduce diversity.
The variational autoencoder was proposed by Kingma and Welling in 2013.
All Facts (6)
Generative molecular models (such as variational autoencoders, diffusion models and reinforcement learning strategies) can directly design new molecules that satisfy specific properties, and then molecular dynamics simulations predict their binding stability with the target protein.
50%UnverifiedA variational autoencoder (VAE) contains an encoder and a decoder, samples latent variables through the reparameterization trick, and its loss function consists of a reconstruction error and a KL divergence regularization term.
50%UnverifiedAmortized inference originates from the design paradigm of variational autoencoders (VAEs), using a shared inference network to reduce the parameter count from O(n) to the network's fixed parameter count, while enabling one-step inference generalization to new observation points.
50%UnverifiedMolecular generation and optimization can leverage graph neural networks (GNN), variational autoencoders (VAE), or diffusion models to generate entirely new molecular structures with specific properties.
50%UnverifiedVariational response generators introduce the idea of variational autoencoders by inserting a latent variable between the encoder and decoder; the latent variable follows a probability distribution and is sampled during generation to introduce diversity.
50%UnverifiedThe variational autoencoder was proposed by Kingma and Welling in 2013.
50%