SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2106.03097 · NeurIPS · 2022

Preservation of the Global Knowledge by Not-True Distillation in Federated Learning

Gihun Lee, Sangmin Bae, Se-Young Yun, Minchan Jeong, Yongjin Shin

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 9 functions out of this paper's own repositories and ran 4 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
Lee-Gihun/FedNTD canonical 1 of 1
TsingZ0/PFL-Non-IID — 2 of 3
thejungwon/gc-fed — 1 of 5
FunctionStatusWhere it lives
Client Ran TsingZ0/PFL-Non-IID/system/flcore/clients/clientntd.py
code served (permissive licence) · get_code("b2630c53b756f58c")
NTD_Loss Ran Lee-Gihun/FedNTD/algorithms/fedntd/criterion.py
code served (permissive licence) · get_code("0f492f116222f383")
NTD_Loss Ran thejungwon/gc-fed/algorithms/fedntd.py
pointer only (licence: NONE) · get_code("95d5cdba6e3175e5")
refine_as_not_true Ran TsingZ0/PFL-Non-IID/system/flcore/clients/clientntd.py
code served (permissive licence) · get_code("b1c10f9b9003aa57")
FedNTD Not yet run thejungwon/gc-fed/algorithms/fedntd.py
pointer only (licence: NONE) · get_code("d862b7d6ee3577ef")
FedNTDClient Not yet run thejungwon/gc-fed/algorithms/fedntd.py
pointer only (licence: NONE) · get_code("04996cc03e29c346")
clientNTD Not yet run TsingZ0/PFL-Non-IID/system/flcore/clients/clientntd.py
code served (permissive licence) · get_code("722bf9f4417d11df")
evaluate Not yet run thejungwon/gc-fed/algorithms/fedntd.py
pointer only (licence: NONE) · get_code("1a0a0b3bfb388c6e")
set_random_seed Not yet run thejungwon/gc-fed/algorithms/fedntd.py
pointer only (licence: NONE) · get_code("55dd9a2df5092e25")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

In federated learning, a strong global model is collaboratively learned by aggregating clients' locally trained models. Although this precludes the need to access clients' data directly, the global model's convergence often suffers from data heterogeneity. This study starts from an analogy to continual learning and suggests that forgetting could be the bottleneck of federated learning. We observe that the global model forgets the knowledge from previous rounds, and the local training induces forgetting the knowledge outside of the local distribution. Based on our findings, we hypothesize that tackling down forgetting will relieve the data heterogeneity problem. To this end, we propose a novel and effective algorithm, Federated Not-True Distillation (FedNTD), which preserves the global perspective on locally available data only for the not-true classes. In the experiments, FedNTD shows state-of-the-art performance on various setups without compromising data privacy or incurring additional communication costs 1 .

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2106.03097")
get_code_for_paper("2106.03097")
have("2106.03097")

Connect an agent — have() is free.