The authors propose MOSEL, a method for a posteriori multi-objective optimization (MOO) in deep neural networks.
Single-pass backpropagation learns Pareto front at cost of standard training
MOSEL reformulates multi-objective optimization as a Stackelberg game, enabling a posteriori Pareto front learning with no extra overhead.
Academic
Elina Rojin Celik · Marcos Medeiros Raimundo · Isabel Valera
Saarland University · Universidade Estadual de Campinas
Research Digest··2 min read
The authors introduce MOSEL, a framework that learns a set of Pareto stationary solutions from a single forward-backward pass.
Why this paper
From Saarland University and Universidade Estadual de Campinas
In one line
A single forward-backward pass can recover a full front of Pareto stationary solutions in multi-objective deep learning.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors (3 noted)
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§