Daniel Francisco Lopes
Federated Learning for Predicting the Next Node in Action Flows
Tese submetida para provas de mestrado em Engenharia
Informática e de Computadores Instituto Superior
Técnico, Universidade de Lisboa.
Abstract
Federated learning is a machine learning approach that allows
different clients to collaboratively train a common model without
sharing their data sets. We focus on centralized federated learning,
where a cen- tral server collects contributions from individual
clients, merges these contributions, and disseminates the results to
all clients. Since clients have different data and classify data
differently, there is a trade- off between the generality of the
common model and the personalization of the classification
results. Current approaches rely on using a combination of a global
model, common to all clients, and multiple local models, that support
personalization. In this work, we report the results of a study, where
we have applied some of these approaches to a concrete use case,
namely the Service Studio platform from OUTSYSTEMS, where Graph Neural
Networks help programmers in the development of
applications. Furthermore, we explore two different approaches which
merge some of the state-of-the-art algorithms so as to develop the
best model for all the different clients. Our results show that one of
the proposed ap- proaches manages to achieve similar performance to
the best-performing algorithms for all the classes of clients and can
even outperform previous algorithms for some classes of clients.
Publicações
- Federated Learning for Predicting the Next Node in Action Flows
- Daniel Francisco Lopes
- MSc Thesis. Instituto Superior Técnico,
Universidade de Lisboa.
- November 2022.
- Available BibTeX, MSC Thesis, and extended abstract, and
mid-term
report.
- Aprendizagem Federada para Previsão do
Próximo Nó em Fluxos de
Ações.
- D. Lopes, J. Nadkarni,
F. Assunção, M. Lopes and L. Rodrigues.
- Actas do
décimo terceiro Simpósio de Informática
(Inforum), Guarda, Portugal, Sep. 2022.
-
- Available BibTeX, extended report (pdf).
- Federated Learning for Predicting the Next Node in
Action Flows
- D. Lopes, J. Nadkarni,
F. Assunção, M. Lopes and L. Rodrigues.
- Accepted
as a Poster in the Workshop on Federated Learning: Recent Advances
and New Challenges (in Conjunction with NeurIPS 2022), New Orleans
(LA,) USA, December 2, 2022.
Luís Rodrigues