Comments (2)
many thanks for the explanation!
from bi-tempered-loss.
In the bi-tempered loss, we don't assume access to the clean labels and instead design a loss function that can handle "possible" noise in the data. This is done by restricting the loss of each example to be below a certain constant value. If your goal is to detect noisy examples in a dataset where you don't have access to clean labels, one possibility is to consider the low-loss approach:
https://arxiv.org/pdf/2104.01493.pdf
https://arxiv.org/pdf/2111.05428.pdf (+ references therein)
The noisy examples (because of their incorrect labels) usually have a higher loss compared to the clean examples (see Figure 1(b) in the first paper). You can use this information to guess which examples are possibly noisy.
from bi-tempered-loss.
Related Issues (15)
- trainning is too slow
- How to calculate "simple integration" in Chapter 3 HOT 1
- Why did you use Bergman divergence instead of KL divergence? HOT 7
- Use sigmod or tempered_sigmoid for prediction? HOT 4
- Nan loss during training HOT 10
- How do I implement Tempered_softmax in Cīŧ HOT 1
- loss_test.py fails in test_gradient_error HOT 1
- Accuracy results on MNIST HOT 3
- Accuracy results on cifar100 HOT 4
- How are the labels corrupted? HOT 2
- Output activation and bi-tempered loss HOT 1
- TF 2.0 Version HOT 2
- why 5 is the default num_iters? HOT 3
- ValueError: Rank mismatch: Rank of labels (received 2) should equal rank of logits minus 1 (received 2) HOT 2
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
đ 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. đđđ
-
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 â¤ī¸ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from bi-tempered-loss.