Libin Zhu
8 papers · 2020–2025 · 4 conferences · across top CS/AI conferences
Achievements
Jump to papers ↓
🌍
Conference Polyglot
(4)
🏃
Academic Marathon
(5)
🌉
Interdisciplinary Bridge
🧭
Keyword Pioneer
🐝
Cross-Pollinator
(15)
Conferences
ICLR (3)
ICML (2)
NIPS (2)
UAI (1)
Top co-authors
Keywords
neural tangent kernel
(3)
transition to linearity
(2)
neural network
(2)
gradient descent
(1)
hessian matrix
(1)
directed acyclic graph
(1)
infinite width
(1)
network initialization
(1)
network width
(1)
feedforward neural network
(1)
eigenvalue bound
(1)
smooth activation function
(1)
lazy training
(1)
deep learning theory
(1)
deep learning
(1)
Papers
Emergence in non-neural models: grokking modular arithmetic via average gradient outer product
ICML 2025
Quadratic models for understanding catapult dynamics of neural networks
ICLR 2024
Catapults in SGD: spikes in the training loss and their impact on generalization through feature learning
ICML 2024
Neural tangent kernel at initialization: linear width suffices
UAI 2023
Restricted Strong Convexity of Deep Learning Models with Smooth Activations
ICLR 2023
Transition to Linearity of Wide Neural Networks is an Emerging Property of Assembling Weak Models
ICLR 2022
Transition to Linearity of General Neural Networks with Directed Acyclic Graph Architecture
NIPS 2022
On the linearity of large non-linear models: when and why the tangent kernel is constant
NIPS 2020