Libin Zhu

Publication Activity (10 Years)

Years Active: 2020-2024
Publications (10 Years): 15

Top Topics

Equivalence Class

Directed Acyclic Graph

Top Venues

Publications

Libin Zhu, Chaoyue Liu, Adityanarayanan Radhakrishnan, Mikhail Belkin
Quadratic models for understanding catapult dynamics of neural networks. ICLR (2024)
Neil Mallinar, Daniel Beaglehole, Libin Zhu, Adityanarayanan Radhakrishnan, Parthe Pandit, Mikhail Belkin
Emergence in non-neural models: grokking modular arithmetic via average gradient outer product. CoRR (2024)
Arindam Banerjee, Pedro Cisneros-Velarde, Libin Zhu, Misha Belkin
Restricted Strong Convexity of Deep Learning Models with Smooth Activations. ICLR (2023)
Arindam Banerjee, Pedro Cisneros-Velarde, Libin Zhu, Mikhail Belkin
Neural tangent kernel at initialization: linear width suffices. UAI (2023)
Libin Zhu, Chaoyue Liu, Adityanarayanan Radhakrishnan, Mikhail Belkin
Catapults in SGD: spikes in the training loss and their impact on generalization through feature learning. CoRR (2023)
Libin Zhu, Parthe Pandit, Mikhail Belkin
A note on Linear Bottleneck networks and their Transition to Multilinearity. CoRR (2022)
Chaoyue Liu, Libin Zhu, Misha Belkin
Transition to Linearity of Wide Neural Networks is an Emerging Property of Assembling Weak Models. ICLR (2022)
Libin Zhu, Chaoyue Liu, Mikhail Belkin
Transition to Linearity of General Neural Networks with Directed Acyclic Graph Architecture. CoRR (2022)
Chaoyue Liu, Libin Zhu, Mikhail Belkin
Transition to Linearity of Wide Neural Networks is an Emerging Property of Assembling Weak Models. CoRR (2022)
Arindam Banerjee, Pedro Cisneros-Velarde, Libin Zhu, Mikhail Belkin
Restricted Strong Convexity of Deep Learning Models with Smooth Activations. CoRR (2022)
Libin Zhu, Chaoyue Liu, Misha Belkin
Transition to Linearity of General Neural Networks with Directed Acyclic Graph Architecture. NeurIPS (2022)
Libin Zhu, Chaoyue Liu, Adityanarayanan Radhakrishnan, Mikhail Belkin
Quadratic models for understanding neural network dynamics. CoRR (2022)
Chaoyue Liu, Libin Zhu, Mikhail Belkin
On the linearity of large non-linear models: when and why the tangent kernel is constant. CoRR (2020)
Chaoyue Liu, Libin Zhu, Mikhail Belkin
Toward a theory of optimization for over-parameterized systems of non-linear equations: the lessons of deep learning. CoRR (2020)
Chaoyue Liu, Libin Zhu, Mikhail Belkin
On the linearity of large non-linear models: when and why the tangent kernel is constant. NeurIPS (2020)