Reputation: 1
Federated learning
I have two kinds of covid 19 datasets as two clients (the first one ST Scan images and the second is XRays images) and I use federated learning approch. The question is can I use those dataset as two clients although they have different type to acheive the heterogeneity and try to solve it ?
Upvotes: 0
Views: 205
Answers (1)
Reputation: 251
Generally speaking when it comes to heterogeneity we think directly to use some personalization methods, you can use a neural network for your federated learning approach and when it comes to retrain your model locally try to find some layers which are user specific (lower layer for specific features for example) keep them on device do not send them for updating the global model. in this case just the upper layer will be trained collaboratively :)
Upvotes: 1
Related Questions
- When switching from federated to centralized dataset -Error may indicate that you're trying to pass a Tensor to a NumPy call
- How to create federated dataset from a CSV file?
- splitting the data into training and testing in federated learning
- Federated Learning for Image Classification in colab
- Federated reinforcement learning
- Implement data generator in federated training
- Create a custom federated data set in TensorFlow Federated
- TFF Federated Learning, Evaluation Approach
- Are there any way to do federated learning with real multiple machines using tensorflow-federated API?
- How to create your own federated dataset and do learning over multiple devices with TensorflowFederated?