Organized Digital Intelligent Network (O.D.I.N)
O.D.I.N is a framework for building "Organized Digital Intelligent Networks".
End-to-end design, versatile, plug-n-play, minimized repetitive work
This repo contains the most comprehensive implementation of variational autoencoder and disentangled representation benchmark.
from odin.fuel import MNIST
from odin.networks import get_networks
from odin.bay.vi import VariationalAutoencoder
ds = MNIST()
train = ds.create_dataset(partition='train')
# optimized architectures for MNIST
networks = get_networks(ds, is_hierarchical=False, is_semi_supervised=False)
# create the VAE
vae = VariationalAutoencoder(**networks)
vae.build(ds.full_shape)
vae.fit(train, max_iter=10000)- VAE
- Hierachical VAE
- Semi-supervised VAE
- Disentanglement Gym
- Faster Classical ML (automatically select GPU implementation)
| Model | Reference/Description | Implementation |
|---|---|---|
|
(Kingma et al. 2014). "Auto-Encoding Variational Bayes" [Paper] | [Code][Example] |
|
(Higgins et al. 2016). "beta-VAE: Learning Basic Visual Concepts with a Constrained Variational Framework" [Paper] | [Code][Example] |
|
Customized version of Beta-VAE, support re-weighing both reconstruction and regularization \(\mathrm{ELBO}=\gamma \cdot E_q[log p(x|z)] - \beta \cdot KL(q(z|x)||p(z|x))\) |
[Code][Example] |
|
(Sønderby et al. 2016) "Ladder variational autoencoder" | [Code][Example] |
|
(Fu et al. 2019) "Cyclical Annealing Schedule: A Simple Approach to Mitigating KL Vanishing" | [Code][Example] |
|
(Chen et al. 2019) "Isolating Sources of Disentanglement in Variational Autoencoders" (regularize the latents' Total Correlation) | [Code][Example] |
|
(Burgess et al. 2018) "Understanding disentangling in beta-VAE" | [Code][Example] |
|
(Kim et al. 2018) "Disentangling by Factorising" | [Code][Example] |
|
(Maaløe et al. 2016) "Auxiliary Deep Generative Models" | [Code][Example] |
|
(Davidson et al. 2018) "Hyperspherical Variational Auto-Encoders" | [Code][Example] |
|
(De Cao et al. 2020) "The Power Spherical distribution" | [Code][Example] |
|
(Kumar et al. 2018) "Variational Inference of Disentangled Latent Concepts from Unlabeled Observations" (I - only_mean=True; II - only_mean=False) | [Code][Example] |
|
(Zhao et al. 2018) "infoVAE: Balancing Learning and Inference in Variational Autoencoders" | [Code][Example] |
|
(Ducau et al. 2017) "Mutual Information in Variational Autoencoders" (max Mutual Information I(X;Z)) | [Code][Example] |
|
(Jing et al. 2020) "Implicit Rank-Minimizing Autoencoder" (Implicit Rank Minimizer) | [Code][Example] |
|
(Figurnov et al. 2018) "Implicit Reparameterization Gradients" (Amortized Latent Dirichlet Allocation - VAE with Dirichlet latents for topic modeling) | [Code][Example] |
|
(Dai et al. 2019) "Diagnosing and Enhancing VAE Models" | [Code][Example] |
|
(Tomczak et al. 2018) "VAE with a VampPrior" | [Code][Example] |
|
(Oord et al. 2017) "Neural Discrete Representation Learning" | [Code][Example] |
| Model | Reference/Description | Implementation |
|---|---|---|
|
(Sønderby et al. 2016) "Ladder variational autoencoder" | [Code][Example] |
|
(Kingma et al. 2016) "Improved variational inference with inverse autoregressive flow" (Bidirectional inference hierarchical VAE) | [Code][Example] |
|
(Zhao et al. 2017) "Learning Hierarchical Features from Generative Models" (Multiple latents connects encoder-decoder from bottom to top in parallel) | [Code][Example] |
| Model | Reference/Description | Implementation |
|---|---|---|
|
Same as FactorVAE, but the discriminator also estimate the density of the labels and unlabeled data (like in semi-GAN) | [Code][Example] |
|
VAE has multiple decoders for different tasks | [Code][Example] |
|
VAE has multiple tasks directly constrain the latents | [Code][Example] |
|
(Kingma et al. 2014) "Semi-supervised learning with deep generative models" [Paper] | [Code][Example] |
|
(Joy et al. 2021) "Capturing label characteristics in VAEs" [Paper] | [Code][Example] |
|
(Trung et al. 2021) "The transitive information theory and its application to deep generative models" [Paper] | [Code][Example] |
DisentanglementGym: fast API for benchmarks on popular datasets and renowned disentanglement metrics.
Dataset support: ['shapes3d', 'dsprites', 'celeba', 'fashionmnist', 'mnist', 'cifar10', 'cifar100', 'svhn', 'cortex', 'pbmc', 'halfmoons']
Metrics support:
- Correlation: 'spearman', 'pearson', 'lasso'
- BetaVAE score
- FactorVAE score
- Mutual Information Estimated
- MIG (Mutual Information Gap)
- SAP (Separated Attribute Prediction)
- RDS (relative disentanglement strength)
- DCI (Disentanglement, Completeness, Informativeness)
- FID (Frechet Inception Distance)
- Total Correlation
- Clustering scores: Adjusted Rand Index, Adjusted Mutual Info, Normalized Mutual Info, Silhouette score.
Automatically accelerated by RAPIDS.ai (i.e. automatically select GPU implementation if available)
- t-SNE [Code]
- UMAP [Code]
- PCA, Probabilistic PCA, Supervised Probabilistic PCA, MiniBatch PCA, Randomize PCA [Code]
- Probabilistic Linear Discriminant Analysis (PLDA) [Code]
- iVector (GPU acclerated) [Code]
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent