Model Function Based Analysis
-
NEW
Global Convergence of Model Function Based Bregman Proximal Minimization Algorithms joint work with
Jalal Fadili Peter Ochs is online at arXiv.
Deep Linear Neural Networks or Deep Matrix Factorization
- Bregman Proximal Framework for Deep Linear Neural Networks joint work with Felix Westerkamp, Emanuel Laude, Daniel Cremers, Peter Ochs is online at arXiv.
Variant of the above paper "Bregman Proximal Gradient Algorithms for Deep Matrix Factorization" which is joint work with Felix Westerkamp, Emanuel Laude, Daniel Cremers, Peter Ochs is accepted at SSVM 2021.
Inertial Bregman Proximal Gradient Algorithms in Non-Convex Optimization
- Beyond Alternating Updates for Matrix Factorization with Inertial Bregman Proximal Gradient Algorithms joint work with Peter Ochs is accepted at NeurIPS 2019.
- Convex-Concave Backtracking for Inertial Bregman Proximal Gradient Algorithms in Non-Convex Optimization joint work with Peter Ochs, Thomas Pock and Shoham Sabach is accepted to SIAM Journal on Mathematics of Data Science (SIMODS).
Deep Learning Theory
- On the Loss Landscape of a class of Deep neural networks with no bad local valleys joint work with Quynh Nguyen and Matthias Hein is accepted at International Conference on Learning Representations (ICLR) 2019.
- Neural Networks Should Be Wide Enough to Learn Disconnected Decision Regions joint work with Quynh Nguyen and Matthias Hein is accepted at International Conference on Machine Learning (ICML) 2018.
Stochastic/Online Optimization for Deep Learning
- Variants of RMSProp and Adagrad with Logarithmic Regret Bounds joint work with
Matthias Hein is accepted at International Conference on Machine Learning (ICML) 2017.
Also see long version containing all proofs here