v0.2.3

GenerativeModels v0.2.3

Diff since v0.2.2

Merged pull requests:

• CompatHelper: bump compat for "Distributions" to "0.24" (#90) (@github-actions[bot])
• Implement NeuralStatistician (#96) (@masenka31)
• Move to Github Actions (#97) (@nmheim)
• Update TagBot (#98) (@nmheim)

v0.2.2

GenerativeModels v0.2.2

Diff since v0.2.1

Merged pull requests:

• Also remove Distances.jl from Project.toml.... (#94) (@nmheim)

v0.2.1

GenerativeModels v0.2.1

Diff since v0.2.0

Merged pull requests:

• Remove Distances.jl dependency (#92) (@nmheim)

v0.2.0

GenerativeModels v0.2.0

Diff since v0.1.1

Closed issues:

• Move saveload.jl to GMExtensions.jl (#88)

Merged pull requests:

• VAMP (#82) (@vitskvara)
• Move to Conditionaldists v0.4 (#86) (@nmheim)
• Remove save/load functions and Rodent (#89) (@nmheim)

v0.1.1

GenerativeModels v0.1.1

Diff since v0.1.0

Closed issues:

• Register (#78)
• Implement VAMP (#79)

Merged pull requests:

• Move to ConditionalDists v0.2 (#81) (@nmheim)

v0.1.0

GenerativeModels v0.1.0

Closed issues:

• Setup CI (#1)
• Add save/load checkpoint functions to src/utils (#2)
• should AbstractVAE functions be defined in seperate file and not in src/models/vae.jl? (#3)
• Variance return function (former: why does softplus on variance kill my ODE optimization?) (#4)
• Regularization of noise (#5)
• Shoudl we start using 0-dimensional Arrays for scalars? (#6)
• AbstractGM (#7)
• Slow compilation/loading (#8)
• Generalize elbo for AbstractVAE (#9)
• Functions for xsize and zsize for AbstractVAE (#10)
• AbstractVAE methods (#11)
• Should VAE have the field prior? (#12)
• Stub methods for mean/variance/mean_var in abstract_vae (#13)
• Think about usage of dropdims (#14)
• Use loglikelihood functions from vita's branch (#15)
• Include xvar field as parametric type in VAE (#16)
• Seperate type for VAE with fixed but learned variance (#17)
• Name variance sigma2 (#18)
• Include constants in likelihoods and KL divergences (#19)
• use mean instead of sum(x)/N (#20)
• New Interface (#21)
• xdim or xsize? (#22)
• Performance (#25)
• generalize ODE decoder definition (#28)
• Clean up (#29)
• Batch in rand as a kwarg? (#30)
• softplus_save does not work with CuArray? (#33)
• CPDF constructors and Conv (#35)
• loss_back_update! and other utils broken with Zygote (#37)
• Zygote and DiffEqFlux (#38)
• xlength and zlength check (#39)
• VAE allows only CGaussian (#40)
• default mean should not be from mean_var (#41)
• unify mmd (#42)
• N-dimensional time series (#43)
• Moving model to gpu ignores NoGradArray (#44)
• Fix GAN (#46)
• update readme (#47)
• GPU issues with NoGradCuArray (#49)
• Zygote cannot repeat? (#59)
• Remove obsolete nogradarray.jl file (#65)
• Prevent zero variance (#66)
• Clean up mmd (#67)
• ImpIement convolutional VAE. (#68)
• fix sign of elbo (#69)
• Type stability (#70)
• Broken gpu macro (#72)
• Do a proper vanilla VAE in tests (#73)

Merged pull requests:

• WIP: Implement new interface (#24) (@nmheim)
• Wasserstein (#27) (@vitskvara)
• GAN (#31) (@vitskvara)
• WIP: Make PDFs GPU ready / Use outer constuctors in PDFs (#32) (@nmheim)
• WIP: Move to Julia 1.3 and Zygote (#36) (@nmheim)
• make untrainable arrays work on GPU (#48) (@nmheim)
• Flux model ODE decoder (#51) (@nmheim)
• ConstSpecGaussian (#56) (@nmheim)
• WIP: ConditionalDists.jl and IMPeasures.jl (#62) (@nmheim)
• Actually move pdfs to ConditionalDists.jl and kl_divergence to IMPeasures.jl ... (#63) (@nmheim)
• fixed gan and other updates (#64) (@vitskvara)
• Type stability (#71) (@nmheim)
• fix broken tests (#74) (@nmheim)
• Update docs / fix elbo (#75) (@nmheim)
• vanilla vae (#76) (@vitskvara)
• bump test coverage (#77) (@nmheim)