10 citations · 10 across the 3 of their papers we have counts for
3 papers
stat.ML2024
Corridor Geometry in Gradient-Based Optimization
Benoit Dherin, Mihaela Rosca
We characterize regions of a loss surface as corridors when the continuous curves of steepest descent -- the solutions of the gradient flow -- become straight lines. We show that c…
stat.ML2023
On discretisation drift and smoothness regularisation in neural network training
Mihaela Claudia Rosca
The deep learning recipe of casting real-world problems as mathematical optimisation and tackling the optimisation by training deep neural networks using gradient-based optimisatio…
cs.LG2022★ 10 cited
Why neural networks find simple solutions: the many regularizers of geometric complexity
Benoit Dherin, Michael Munn, Mihaela Rosca +1
In many contexts, simpler models are preferable to more complex models and the control of this model complexity is the goal for many methods in machine learning such as regularizat…