Comments (5)
Hi,
In the same veins; I have some questions that I couldn't answer from only reading the code. In the provided implementation, is the avoidance of empty clusters by reassigning them is done automatically by faiss
clustering? Another question if you may, between each two epochs, k-means is run and the new labels are reassigned, so from epoch to epoch the significance of each class label changes ? is this a normal behavior and helps the model in a self supervised way or is there a trick to have a consistency of the labels across epochs (like one-to-one assignments).
Thank you very this awesome work.
from deepcluster.
Hi,
Thanks for your interest!
What you can do is sampling the batches based on a uniform distribution over the clusters (see here. You may also split one cluster in two whenever you get an empty cluster or add an entropic term in your k-means loss.
from deepcluster.
Hi,
Thanks for your interest and sorry for the delay of my reply.
Nothing prevents empty clusters from occurring in my implementation. This is not done automatically by faiss.
The significance of each class labels changes from a clustering assignment to another, that's why we reset the last fully connected layer. It is indeed an expected behavior.
Please re-open the issue if you have further questions.
from deepcluster.
Thanks, I did see that in the new paper, deeper clustering, there is indeed the reassignment part.
from deepcluster.
Hi,@yassouali
I have a same problem about the reassignment of the labels between each two epochs. which is the new paper you said, Thank you.
from deepcluster.
Related Issues (20)
- AttributeError: 'Clustering' object has no attribute 'obj' HOT 3
- RuntimeError: invalid argument 5: k not in range for dimension at /pytorch/aten/src/THC/generic/THCTensorTopK.cu:23 HOT 2
- train dataset that is not sorted in different folders HOT 1
- Hey, can you please share how you have solved this problem because I am also getting the same type of error which m not able to solve.hope you can help me out. Thank you. HOT 2
- Do I need labels or pseudo-labels as input for clustering? HOT 8
- A problem about the clusters and pseudolabels every epoch HOT 8
- A problem why use different optimizer for top layer and other layers HOT 1
- About the clustering loss HOT 1
- About the clustering loss HOT 3
- Is this model universal on different datasets? HOT 1
- Why does the deep cluster model cluster unevenly when using MNIST datasets? HOT 1
- Dealing with empty clusters HOT 3
- Gradcam
- about NMI
- Could I use it for image color clustering?
- PackagesNotFoundError
- Faiss
- how to avoid empty clusters. HOT 1
- Loading
- The pre-trained AlexNet on ImageNet
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 deepcluster.