Releases: nmichlo/disent
Releases · nmichlo/disent
v0.8.0
v0.7.2
v0.7.1
v0.7.0
v0.6.3
v0.6.2
v0.6.1
v0.6.0
Fixes
- MPI3D was not correctly loaded, first few factors were misaligned
- Recomputed statistics for new datasets and updated configs
Additions
- Added
disent.data.data.DataFileSprites
, a custom version of sprites- Added experiment configs and computed dataset statistics
- Multiple version of
disent.dataset.data.Mpi3dData
now exist, for different use cases because the dataset is so large- added
Mpi3dHdf5Data
-- converts the files to hdf5 to stream from disk, but very slow to load into memory directly - added
Mpi3dNumpyData
-- loads the files directly into memory (quick), cannot read from disk - changed:
Mpi3dData
is now a wrapper around both of the above, and the mode can be specified within_memory
- added
disent.dataset.util.state_space.StateSpace
- Added init checks
- Added helper method
invert_factor_idxs
that returns the unspecified factor indices, or the inverse set. - Added helper method
sample_indices
that samples valid indices in the range of the dataset. - Improved sampling and other methods that take in factors to first call
normalise_factor_idxs
so that we can use factor names in these functions instead. - Added helper method
sample_random_factor_traversal_grid
that samples a grid of traversals, one for each ground-truth factor.
- Added
disent.util.inout.paths.modify_ext(...)
that modifies the extension of a path
Breaking changes
- move
disent.dataset.util.npz
todisent.dataset.util.formats.npz
- move
disent.dataset.util.hdf5
todisent.dataset.util.formats.hdf5
disent.util.inout.hashing.hash_file
now hasmissing_ok=False
by default
Minor Fixes
- Fix
stalefile
now correctly handles missing files - Various plotting fixes, now functions support RGBa images not just grey or RGB images.
New Tests
- Added some new tests for both dataset formats and state spaces
TODO:
- Added
Teapots3dData
but it is not complete, needs to be converted to a "random" dataset, as this dataset does not actually have valid ground truth factors in the form of a state space, rather they are randomly sampled.
v0.5.1
v0.5.0
This release marks the end of my MSc. and splitting the research out into its own repository!
- The repo was previously setup such that development took place on an
xdev
branch. An automated script was then used to clean this branch of research code and commit the changes to thedev
branch, which was then published. - This has now been disabled in favour of standard dev practice. I no longer need to maintain the old research code and can incorporate this functionality directly into disent.
MSc. Additions
-
disent.dataset.data
- Various new datasets!XYObjectData
andXYObjectShadedData
equivalent datasets with different representations of their ground-truth factors. Disentanglement performance is affected by the choice of ground-truth factors even if the data is exactly the same!XYSquaresData
is an adversarial dataset for VAEs that use pixel-wise reconstruction losses. VAEs usually perform terribly on this dataset in terms of disentanglement performance. This dataset contains three squares that can move across a non-overlapping grid.XYSingleSquareData
is likeXYSquaresData
but only has a single square that can move across the image.XColumnsData
is a simplistic version ofXYSquaresData
that is still adversarial, but only moves columns left and right instead of an object across a grid.
-
disent.frameworks.vae
AdaNegTripletVae
aka. "ada_tvae": Supervised disentanglement framework that uses our proposed Adaptive Triplet Loss to disentangle representations and introduce axis-alignment. Triplets are constructed using the L1 distance between ground-truth factors.DataOverlapTripletVae
aka. "ada_tvae_d": Unsupervised version of theAdaNegTripletVae
that order triplets using the distances between datapoints in terms of the reconstruction loss. Distances within disentanglement datasets often correspond to the distances between ground-truth factors, suggesting disentanglement is accidental!
-
disent.frameworks.ae
AdaNegTripletAe
aka. "ada_tae" - The AE version ofAdaNegTripletVae
DataOverlapTripletAe
aka. "ada_tae_d" - The AE version ofDataOverlapTripletVae
AdaAe
- The AE version of theAdaVae
-
disent.metrics
flatness_components
consists of three separate metricsdistances
: measure the rank correlation between ground-truth distances and latent distanceslinearity
: measure how well factor traversal embeddings lie on an arbitrarily rotated n-dimensional lineaxis-alignment
: measure how well factor traversal embeddings correspond to a single latent variable (ie. an n-dimensional line that is axis-aligned).
flatness
an older metric that measures the path length of factor traversal embeddings over the max distance between points.
-
experiment/configs
updated to included configs for all the added classes, frameworks, datasets, metrics and features!- new schedules
schedule/adanegtvae_*.yaml
that should be used with the Adaptive Triplet frameworks. Otherwise these frameworks do not learn.
- new schedules
MSc. Removals
- All the remaining research code contained in
research/*
has been deleted
Add Examples
- Added an example
docs/examples/extend_experiment
of how to override or extend the disent experiment conifigs! This is useful for creating your own research! - Added an example of plotting various aspects of disent
docs/examples/plotting_examples
.
Fixes
- Fixed tests for new locations
- Added appropriate entries to the registry