Ditto

KO | EN

gitlite — search

#hacktoberfest #llm #android #machine-learning #rust #deep-learning #linux #docker #macos #awesome #vue #awesome-list #security #ios

(★ 55)

No description available.

파일 탐색기

최종 버전 다운로드 (.zip)

- - baseline_constants.cpython-36.pyc
  - client.cpython-36.pyc
  - model.cpython-36.pyc
  - server.cpython-36.pyc
- - __init__.py
  - sys_metrics.csv
  - visualization_utils.py
  - writer.py
- - erm_cnn_log_reg.py
- - __init__.py
  - constants.py
  - model_utils.py
  - so_tag_utils.py
  - torch_utils.py
- .DS_Store
- baseline_constants.py
- client.py
- main.py
- model.py
- run.sh
- server.py
README.md

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Ditto: Fair and Robust Federated Learning Through Personalization

This repository contains the code and experiments for the manuscript:

> [Ditto: Fair and Robust Federated Learning Through Personalization](https://arxiv.org/abs/2012.04221)
>

Fairness and robustness are two important concerns for federated learning systems.
In this work, we identify that *robustness* to data and model poisoning attacks and *fairness*, measured as the uniformity of performance across devices, are competing constraints in statistically heterogeneous networks. 
To address these constraints, we propose employing a simple, general framework for personalized federated learning, Ditto, and develop a scalable solver for it. 
Theoretically, we  analyze the ability of Ditto to achieve
fairness and robustness simultaneously on a class of linear problems.
Empirically, across a suite of federated datasets, we show that Ditto not only achieves competitive performance relative to recent personalization methods, but also enables more accurate, robust, and fair models relative to state-of-the-art fair or robust baselines.

This pytorch implementation is based off of the code from [Simplicial-FL repository](https://github.com/krishnap25/simplicial-fl) ([Laguel et al. 2021](https://ieeexplore.ieee.org/document/9400318)).

### *We also provide [Tensorflow implementation](https://github.com/litian96/ditto)*

## Preparation

### Downloading dependencies

```
pip3 install -r requirements.txt
```

## Run on federated benchmarks

(A subset of) Options in `models/run.sh`:

* `dataset` chosen from `[so]`, where so is short for StackOverflow.
* `aggregation` chosen from `['mean', 'median', 'krum']`.
* `attack` chosen from `['label_poison', 'random', 'model_replacement']`. 
* `num_mali_devices` is the number of malicious devices. 
* `personalized` indicates whether we want to train personalized models.
* `clipping` indicates whether we want to clip the model updates while training the global model.
* `k_aggregator` indicates whether we want to run k-loss/k-norm.

Ditto

파일 탐색기

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Ditto

파일 탐색기

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README