Convergence under Lipschitz smoothness of ease-controlled Random Reshuffling gradient Algorithms

1 code implementation • 4 Dec 2022 • Giampaolo Liuzzi, Laura Palagi, Ruggiero Seccia

We aim to define ease-controlled modifications of the IG/RR schemes, which require a light additional computational effort and can be proved to converge under very weak and standard assumptions.

Block Layer Decomposition schemes for training Deep Neural Networks

no code implementations • 18 Mar 2020 • Laura Palagi, Ruggiero Seccia

Deep Feedforward Neural Networks' (DFNNs) weights estimation relies on the solution of a very large nonconvex optimization problem that may have many local (no global) minimizers, saddle points and large plateaus.

A gray-box approach for curriculum learning

no code implementations • 17 Jun 2019 • Francesco Foglino, Matteo Leonetti, Simone Sagratella, Ruggiero Seccia

Curriculum learning is often employed in deep reinforcement learning to let the agent progress more quickly towards better behaviors.

reinforcement-learning Reinforcement Learning (RL) +1