Urmăriți
Daniel Kunin
Daniel Kunin
Adresă de e-mail confirmată pe stanford.edu - Pagina de pornire
Titlu
Citat de
Citat de
Anul
Pruning neural networks without any data by iteratively conserving synaptic flow
H Tanaka*, D Kunin*, DL Yamins, S Ganguli
Neural Information Processing Systems (NeurIPS), 2020
7072020
Loss landscapes of regularized linear autoencoders
D Kunin, J Bloom, A Goeva, C Seed
International Conference on Machine Learning (ICML) Oral, 2019
1062019
Neural mechanics: Symmetry and broken conservation laws in deep learning dynamics
D Kunin, J Sagastuy-Brena, S Ganguli, DLK Yamins, H Tanaka
International Conference on Learning Representations (ICLR), 2020
782020
Beyond the Quadratic Approximation: the Multiscale Structure of Neural Network Loss Landscapes
C Ma, D Kunin, L Wu, L Ying
Journal of Machine Learning Research, 2022
58*2022
Two routes to scalable credit assignment without weight symmetry
D Kunin, A Nayebi, J Sagastuy-Brena, S Ganguli, J Bloom, D Yamins
International Conference on Machine Learning (ICML), 2020
392020
Noether’s Learning Dynamics: Role of Symmetry Breaking in Neural Networks
H Tanaka, D Kunin
Neural Information Processing Systems (NeurIPS), 2021
34*2021
The asymmetric maximum margin bias of quasi-homogeneous neural networks
D Kunin, A Yamamura, C Ma, S Ganguli
International Conference on Learning Representations (ICLR) Spotlight, 2022
252022
Stochastic collapse: How gradient noise attracts sgd dynamics towards simpler subnetworks
F Chen*, D Kunin*, A Yamamura*, S Ganguli
Journal of Statistical Mechanics: Theory and Experiment, 2024
242024
The Limiting Dynamics of SGD: Modified Loss, Phase-Space Oscillations, and Anomalous Diffusion
D Kunin, J Sagastuy-Brena, L Gillespie, E Margalit, H Tanaka, S Ganguli, ...
Neural Computation 36 (1), 151-174, 2023
20*2023
Get rich quick: exact solutions reveal how unbalanced initializations promote rapid feature learning
D Kunin, A Raventós, C Dominé, F Chen, D Klindt, A Saxe, S Ganguli
Neural Information Processing Systems (NeurIPS) Spotlight, 2024
62024
From Lazy to Rich: Exact Learning Dynamics in Deep Linear Networks
CCJ Dominé, N Anguita, AM Proca, L Braun, D Kunin, PAM Mediano, ...
arXiv preprint arXiv:2409.14623, 2024
12024
A Quasistatic Derivation of Optimization Algorithms' Exploration on Minima Manifolds
C Ma, D Kunin, L Ying
Sistemul nu poate realiza operația în acest moment. Încercați din nou mai târziu.
Articole 1–12