activity
20202023
most citedScatterbrain: Unifying Sparse and Low-rank Attention Approximation

9 citations · 49 across the 15 of their papers we have counts for

collaborators
Showing 2021Show all

8 papers · 1 filter

cs.LG2021

Fast Graph Neural Tangent Kernel via Kronecker Sketching

Shunhua Jiang, Yunze Man, Zhao Song +2

Many deep learning tasks have to deal with graphs (e.g., protein structures, social networks, source code abstract syntax trees). Due to the importance of these tasks, people turne…

cs.LG2021

Training Multi-Layer Over-Parametrized Neural Network in Subquadratic Time

Zhao Song, Lichen Zhang, Ruizhe Zhang

We consider the problem of training a multi-layer over-parametrized neural network to minimize the empirical risk induced by a loss function. In the typical setting of over-paramet…

cs.LG2021★ 7 cited

Pixelated Butterfly: Simple and Efficient Sparse training for Neural Network Models

Tri Dao, Beidi Chen, Kaizhao Liang +4

Overparameterized neural networks generalize well but are expensive to train. Ideally, one would like to reduce their computational cost while retaining their generalization benefi…

cs.LG2021★ 3 cited

Breaking the Linear Iteration Cost Barrier for Some Well-known Conditional Gradient Methods Using MaxIP Data-structures

Anshumali Shrivastava, Zhao Song, Zhaozhuo Xu

Conditional gradient methods (CGM) are widely used in modern machine learning. CGM's overall running time usually consists of two parts: the number of iterations and the cost of ea…

cs.LG2021

Online MAP Inference and Learning for Nonsymmetric Determinantal Point Processes

Aravind Reddy, Ryan A. Rossi, Zhao Song +6

In this paper, we introduce the online and streaming MAP inference and learning problems for Non-symmetric Determinantal Point Processes (NDPPs) where data points arrive in an arbi…

cs.LG2021★ 9 cited

Scatterbrain: Unifying Sparse and Low-rank Attention Approximation

Beidi Chen, Tri Dao, Eric Winsor +3

Recent advances in efficient Transformers have exploited either the sparsity or low-rank properties of attention matrices to reduce the computational and memory bottlenecks of mode…