Comments (4)
Hi, thanks for your attention! I forgot to delete the useless settings, sorry for the misleading. I have tried several cases for the second part (CPL) of our work, i.e., 'normal', 'soft', 'vote', 'vote_threshold', 'vote_soft', and 'no', please refer to utils.py for more details about the implementation of each case.
'no': we use all predictions from the other branch as the supervision to supervise the prediction logits of the selected branch.
'normal': we only use confident predictions as the supervision by comparing the prediction probability with the threshold.
'soft': we use confident predictions as the supervision with a co-efficient w1, and we also use unconfident predictions as the supervision with a co-efficient w2.
'vote': if two branches have the same predictions, we set the no-conflicting predictions as the supervision with a co-efficient w1, and we also use conflicting predictions as the supervision with a co-efficient w2.
'vote_threshold': we select conflicting but confident predictions and use the predictions as the supervision with a co-efficient w1, and we use other predictions as the supervision with a co-efficient w2.
'vote_soft': can be treated as the combination of 'vote_threshold' and 'soft', we select no-conflicting predictions as the supervision with a co-efficient w0 (which is set as 1 in the code), conflicting but confident predictions as the supervision with a co-efficient w1 and conflicting but unconfident predictions as the supervision with a co-efficient w2.
In one word, you just need to set mode_confident as 'vote_threshold'.
from ccvc.
Hi! As introduced in the paper, the final version is "vote_soft", and is the co-efficient w0=w2=1, w1=wc=2?
from ccvc.
Hi, please refer to train.sh, we used vote_threshold.
Also refer to utils.py for details of vote_threshold and vote_soft.
We kept the codes as we did such experiments, maybe you can run different strategies and compare the performance. Based on our experiments, vote_threshold is better than vote_soft.
from ccvc.
ohohoh, my mistake. So sorry! Thank you!
from ccvc.
Related Issues (14)
- Results HOT 2
- Confusion about Equations HOT 3
- Questions about retraining HOT 4
- Question about the inplanes of ResNet HOT 1
- Question about the setting in your experiment HOT 4
- Question about the GPU HOT 1
- About Supervised loss and consistency loss HOT 11
- Question about the evaluation method HOT 1
- About unlabeled.txt HOT 3
- Question about the CPL loss HOT 2
- 关于子网坍缩 HOT 3
- uestion about the setting of "mode_confident"
- L2 normalization和CosineSimilarity
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from ccvc.