Skip to content

quinngroup/Generative-Models

Repository files navigation

README

Repository for the JWZ group.

1 A tutorial resource for VAEs emphasizing the mathematical foundation

2 A python package for summarizing your models similar to the Keras .summary() method. Download the package into your main project wd e.g

--Src
  --Models
    myClass.py
    --pytorch-summary-master

3 VAE with a modified mixture of variationals prior (VampPrior)

4 VAE with a modified mixture of variationals prior (VampPrior) github page

5 Beta-VAE paper. Also discussing the notions of disentanglement

6 InfoVAE paper covering maximization of mutual information between inputs and intermediate codes a la InfoGAN techniques.

7 "A General and Adaptive Robust Loss Function." Worth a read. Seems like a very interesting dynamic loss function to use. It can be altered and scheduled to smoothly interpolate between various loss functions which occur as special cases of this one.

8 A very good explanation of computational graphs.

9 Github repository for generating moving digits. We will use these scripts to generate train/test data for our video based generative models. We will be able to control key independent generative features and hence be able to see if our model truly captures critical features.

10 Basic VAE example in PyTorch

11 Basic intro to Transpose Conv

12 Basic intro to Transpose Conv

13 Basic intro to Conv arithmatic of all forms

14 Introduction to data topology

15 LSTM tutorial

16 Spectral clustering tutorial

17 Eigen-cuts algorithm

18 Utility package we ought to implement

19 Two-stream dynamic texture synthesis

20 MONet: Unsupervised scene decomposition and representation

21 Spatial Broadcast Decoder

22 IODINE Network

23 Augmented MNIST data set, including texture-in-texture generation

24 Further explanation for VAE objective functino, ELBO and loss

25 Paper on using compressed representation of video data to truncate noise from low-frequency motion

26 Paper exploring the idea and implementation of general homemorphic manifolds extending the standard guassian prior

27 Paper creating an model which operates on clustering/classifying sets as objects rather than vectors

28 Paper on the equivariance of models (through parameter sharing) using resistance to group actions

29 "PROBABILISTIC SYMMETRY AND INVARIANT NEURAL NETWORKS"

30 "Deformable Convolutional Networks"

31 "ON LARGE-BATCH TRAINING FOR DEEP LEARNING: GENERALIZATION GAP AND SHARP MINIMA"

32 Paper exploring (fast) time-series clustering

33 B3DF paper

34 MonNet architecture, which is a good look at iterative refinmenet in models, along with a great introduction to the utility of masks

35 IODINE network architecture, which is an evolutoin of

36 High-fidelity reconstruction using discretized encoding vectors