Winter 2022-2023 (Dec - Feb)

[01/09/2023] Game Theoretic Perspective of Loss Landscapes & Implicit Self-Regularization of Deep Learning

A game-theoretic perspective of deep neural networks (Ren et al., Theor. Comput. Sci. 2023)

Presenter (offline): Laura Schmid (Postdoc at OSI Lab @ KAIST AI)
Slide and Paper
(in English)

Implicit Self-Regularization in Deep Neural Networks: Evidence from Random Matrix Theory and Implications for Learning (Martin & Mahoney, JMLR 2021)

Presenter (offline): Junghyun Lee (OSI Lab & OptiML Lab @ KAIST AI)
Slide and Paper

[12/12/2022] High-dimensional online learning algorithms & loss landscapes

Exact Asymptotics of High-dimensional Online Learning Algorithms

Presenter (offline): Hojung Jung (OSI Lab @KAIST AI)
Slide
Papers presented:
- Online learning for sparse PCA in high dimensions: Exact dynamics and phase transitions (Wang & Lu, ITW 2016)
- Scaling Limit: Exact and Tractable Analysis of Online Learning Algorithms with Applications to Regularized Regression and PCA (Wang et al., arXiv 2017)
- The scaling limit of high-dimensional online independent component analysis (Wang & Lu, J. Stat. Mech. 2019)

Loss landscapes and optimization in over-parametrized non-linear systems and neural networks (Liu et al., Appl. Comput. Harmon. Ana. 2022)

Presenter (offline): Hanseul Cho (OptiML Lab @KAIST AI)
Slide and Paper

Page updated

Google Sites

Report abuse