iqua/flsim

A simulation framework for Federated Learning written in PyTorch

flsim's Issues

DQN Code Release

Hi there, I have a quick question, is there any plan of releasing the DQN code? Would love to evaluate this work and extend it further.

Can you share the complete code

My DQN implementation based on flsim

Hi there,

Thanks for the good work! Just a question, is the DQN model trained (by selecting 1 participating device as mentioned in the paper) before being deployed to the server for FL communications? Or does it trained in parallel with the FL communications? If so, how? Because in FL communications you are using DQN to select top k devices for training right?

Anyone else can help me understand this? Thanks!

Problem of code

Hello! I'm sorry to bother you. In your paper, you said that you need use the PCA to achieve Dimension Reduction, but I cannot find in your codes. Could you tell me where is this part in your codes? Thank you very much!

Some doubts and questions

I understand that for some reasons you might not have been able to release your complete code but I would highly appreciate if you could help me answering some questions about your implementation.

1. The validation set on server, how much data it has and is it taken from original training (before partitioning) or test set?
2. Do you train your DQN network with one optimization step after each communication round (after pushing the latest experience into replay memory) or multiple steps? Do you wait for the memory to collect some experience or train DQN even with 1 entry? What is the DQN training batch size?
3. what is the optimization algorithm and learning rate used to train the DQN network?
4. What is the frequency of updating the target network (from the learning DQN)?
5. do you use learning rate decay as in FedAvg? Does it match their numbers?
6. Do you use a discounting factor for reward (\gamma in your paper)?

Thank you in advance!

Syntax error

flsim/scripts/pca.py

Line 57 in 59d39e6

logging.into('Writing output to binary...')

It should be logging.info(.) instead of logging.into(.).

about RL in the Paper

Hello Sir,

I have read your paper and I am curious about the time to train an RL agent for MNIST dataset by using 80% no-iid dataset. Do you mind to tell that?

Best wishes

Can you please open source your code of FAVOR?

Can you please open source your code of FAVOR?
The core part of codes is missing(RL agent) and the results couldn't be reproduced.

how to reproduce your experimental results completely

Thank you for your nice codes.

RL code disable

I am curious about the work in your paper, but I found that it is unavailable to the DQN implementation. Could you share the code about server.dqn and server.train?

Not performing PCA - pref is not define

Code is not performing PCA.

flsim/scripts/pca.py

Line 56 in 59d39e6

output = [(clients[i].client_id, clients[i].pref, weight) for i, weight in enumerate(weight_vecs)]

clients[i].pref is not define.

AttributeError: 'Client' object has no attribute 'pref'