ixxi-dante an2vec issues

ixxi-dante / an2vec

Bringing node2vec and word2vec together for cool stuff

GNU General Public License v3.0

22 stars 6 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Encode σ directly in the OrthogonalGaussian codec

#53 wehlutyk closed 6 years ago
4
Dynamics of language use

#52 wehlutyk opened 6 years ago
0
Test model against adversarial attacks

#51 wehlutyk opened 6 years ago
0
Minibatch effect and strategies

#50 wehlutyk opened 6 years ago
0
Test BlogCatalog training in minibatch

#49 wehlutyk closed 5 years ago
3
Test various fixes to blogcatalog training

#48 wehlutyk closed 6 years ago
6
Visualise blogcatalog feature reconstruction to get a quick view of quality

#47 wehlutyk closed 6 years ago
2
Clean up data filenames for blogcalatog exploration

#46 wehlutyk closed 6 years ago
0
[WIP] Reproducing original GAE Cora embedding

#45 wehlutyk closed 6 years ago
0
Reproduce Cora embedding from VGAE paper by Kipf & Welling 2016

#44 wehlutyk closed 6 years ago
2
Train BlogCatalog with no feature reconstruction (and normal input features)

#43 wehlutyk closed 6 years ago
3
[WIP] BlogCatalog exploration

#42 wehlutyk closed 6 years ago
0
Train BlogCatalog with a much higher weight (e.g. 1000) on adjacency reconstruction

#41 wehlutyk closed 6 years ago
2
Train BlogCatalog with no feature reconstruction (and Id input features)

#40 wehlutyk opened 6 years ago
1
Implement Bernoulli codec for feature reconstruction

#39 wehlutyk closed 6 years ago
0
Train BlogCatalog without weighing positive/negative errors in adjacency reconstruction loss

#38 wehlutyk closed 6 years ago
5
Train BlogCatalog without node features

#37 wehlutyk opened 6 years ago
1
Re-train BlogCatalog with embedding dimension 2

#36 wehlutyk closed 6 years ago
3
[WIP] Full batch sensitivity analysis

#35 wehlutyk opened 6 years ago
0
[WIP] Minibatch validation

#34 wehlutyk opened 6 years ago
0
Left-over todos

#33 wehlutyk closed 6 years ago
5
Capturing feature-network dependencies

#32 wehlutyk opened 6 years ago
1
Train BlogCatalog with higher embedding dimension

#31 wehlutyk closed 6 years ago
4
Make more training animations

#30 wehlutyk opened 6 years ago
1
Minibatch sensitivity analysis

#29 wehlutyk closed 6 years ago
13
Minibatch validation: with no GCN, feature-only reconstruction

#28 wehlutyk opened 6 years ago
0
Initial BlogCatalog application

#27 wehlutyk closed 6 years ago
1
Explore deploying on the PSMN (CPU-only) cluster

#26 wehlutyk opened 6 years ago
2
Profile TF to see which ops are slow

#25 wehlutyk opened 6 years ago
2
Use sparse features and keep all the GCN computations sparse

#24 wehlutyk opened 6 years ago
1
Replace `tensordot` call by `while_loop` in `Bilinear` layer, to only compute what we need

#23 wehlutyk closed 5 years ago
4
Mini batch optimisation

#22 wehlutyk closed 6 years ago
0
Mini-batching optimisation

#21 wehlutyk closed 6 years ago
4
Possible model improvements / optimisations

#20 wehlutyk opened 6 years ago
1
Validate mini-batch with convolutionless feature-only reconstruction

#19 wehlutyk closed 6 years ago
9
Mini-batch sampling

#18 wehlutyk closed 6 years ago
1
Set up a validation metric for adjacency and feature reconstruction

#17 wehlutyk opened 6 years ago
1
Decide on a license

#16 wehlutyk opened 6 years ago
4
Mini-batch sampling strategy

#15 wehlutyk closed 6 years ago
5
Decide which adjacency reconstruction normalisation to use

#14 wehlutyk closed 6 years ago
3
Re-evaluate the need for custom keras version

#13 wehlutyk closed 6 years ago
1
Full batch sensitivity analysis

#12 jaklevab closed 6 years ago
3
Use non-recursive importance sampling for mini-batches on dense graphs?

#11 wehlutyk closed 6 years ago
1
[WIP] Mini-batch training in VAE

#10 wehlutyk closed 6 years ago
1
Rate trade-off between feature and network structure

#9 wehlutyk opened 6 years ago
0
Scale up the VAE tests

#8 wehlutyk closed 6 years ago
4
Read "Efficient Backprop" to check for mistakes in the implementation

#7 wehlutyk closed 6 years ago
6
Full batch sensitivity analysis (adjacency reconstruction, feature reconstruction, and both)

#6 wehlutyk opened 6 years ago
6
How much does the trained ξ improve on the untrained ξ?

#5 wehlutyk closed 6 years ago
2
Support saving/loading models in VAE

#4 wehlutyk closed 6 years ago
1