brain-research realistic-ssl-evaluation issues

brain-research / realistic-ssl-evaluation

Open source release of the evaluation benchmark suite described in "Realistic Evaluation of Deep Semi-Supervised Learning Algorithms"

Apache License 2.0

458 stars 98 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

why this implementation of pseudo-labeling

#35 hzhz2020 opened 2 years ago
0
where is data augmentation performed?

#34 hzhz2020 closed 4 years ago
1
Tabular results for Fig.4

#33 taoyang1122 opened 4 years ago
2
Script for downloading and preprocessing CIFAR and svhn dataset

#32 AngranLi closed 4 years ago
1
Why there is no dropout in the model???

#31 wanginwang opened 4 years ago
0
VAT GPU vRAM usage

#30 hanscol closed 4 years ago
4
How is the error rate and uncertainty calculated?

#29 varunnair18 closed 5 years ago
1
Is there a way to download ImageNet weights for WRN-28-2?

#28 varunnair18 closed 5 years ago
8
VAT implementation is wrong

#27 AskAukNuTutor closed 5 years ago
2
Why not use original pseudo-label-implement as baseline?

#26 CheukNgai closed 5 years ago
3
L1/L2 regularization not found in the code

#25 aretor closed 5 years ago
3
Fix tmuxp link in README.md

#24 alexandra-zaharia closed 5 years ago
1
Remove the 'cifar_unnormalized' dataset

#23 avital opened 5 years ago
1
Add 'none' consistency model for fully supervised

#22 craffel closed 5 years ago
0
Add seeding to dataset creation, remove label maps

#21 craffel closed 5 years ago
6
Remove cifar_unnormalized

#20 craffel opened 5 years ago
1
Learning rate for fully supervised baseline is wrong

#19 craffel closed 5 years ago
0
README says to build_tfrecords.py for imagenet_32 before downloading imagenet_32

#18 craffel closed 5 years ago
0
Randomness in build_tfrecords.py + label maps

#17 michalzajac-ml closed 5 years ago
2
cifar_unnormalized should have normalize=False

#16 joschu closed 5 years ago
1
Ensure that summaries are written for one-shot evalutation

#15 gar1t closed 5 years ago
1
Mark global_step_init as used to prevent spurious warning msg

#14 gar1t opened 5 years ago
1
Flag to specify location of summaries for evaluate

#13 gar1t closed 5 years ago
1
Special value 'latest' to evaluate latest available checkpoint

#12 gar1t opened 5 years ago
1
Use -1 as default for n_labeled

#11 gar1t closed 5 years ago
1
Use primary dataset as default for secondary

#10 gar1t closed 5 years ago
1
Results in the paper for low-regime data

#9 TheRevanchist closed 5 years ago
1
consistency_model=mean_teacher, when you do fully supervised learning

#8 TheRevanchist closed 5 years ago
1
add an argument to use only labeled data in training process.

#7 DoctorKey closed 5 years ago
9
It might use unlabeled data to train "fullysup"

#6 DoctorKey opened 5 years ago
2
fix bugs in evaluate_model.py

#5 DoctorKey closed 5 years ago
1
Add run for ImageNet pre-training with fewer classes

#4 avital closed 6 years ago
5
Fix ZCA normalization on ImageNet32

#3 avital closed 6 years ago
1
Why use WRN-28-2?

#2 bl0 closed 6 years ago
1
Code Release

#1 avital closed 6 years ago
21