• Publications
  • Influence
On the Convergence of FedAvg on Non-IID Data
TLDR
We analyze the convergence of \texttt{FedAvg} on non-iid data and establish a convergence rate of $\mathcal{O}(\frac{1}{T})$ for strongly convex and smooth problems, where $T$ is the number of SGDs. Expand
  • 171
  • 36
  • PDF
Improving CUR matrix decomposition and the Nyström approximation via adaptive sampling
TLDR
We establish a general error bound for the adaptive column/row sampling algorithm, based on which we propose more accurate CUR and Nystrom algorithms with expected relative-error bounds. Expand
  • 142
  • 13
  • PDF
Efficient Subspace Segmentation via Quadratic Programming
TLDR
We explore in this paper efficient algorithmic solutions to robust subspace segmentation. Expand
  • 71
  • 10
  • PDF
Nonconvex Relaxation Approaches to Robust Matrix Recovery
TLDR
We propose a nonconvex optimization model for the low-rank matrix recovery problem. Expand
  • 58
  • 10
  • PDF
GIANT: Globally Improved Approximate Newton Method for Distributed Optimization
TLDR
For distributed computing environment, we consider the empirical risk minimization problem and propose a distributed and communication-efficient Newton-type optimization method. Expand
  • 46
  • 10
  • PDF
New calculation methods of diurnal distribution of solar radiation and its interception by canopy over complex terrain
The prescription of diurnal radiation distribution and the consideration of topographic impact on canopy radiation interception are often required in ecological modelling studies. The most commonlyExpand
  • 62
  • 6
Colorization by Matrix Completion
TLDR
This paper proposes a novel approach to the colorization problem by formulating it as a matrix completion problem and resort to an augmented Lagrange multiplier algorithm for solving it. Expand
  • 25
  • 5
Dynamics of surface albedo of a boreal forest and its simulation
Abstract Surface albedo determines the distribution of solar radiation between the earth's surface and the atmosphere. It affects the global climate directly by altering surface energy balance, andExpand
  • 52
  • 5
Scalable Kernel K-Means Clustering with Nystrom Approximation: Relative-Error Bounds
TLDR
Kernel $k$-means clustering with rank-restricted Nystrom approximation can correctly identify and extract a far more varied collection of cluster structures than the linear $k$, but it is not a universal solution to all clustering problems. Expand
  • 41
  • 4
  • PDF
Communication Efficient Decentralized Training with Multiple Local Updates
TLDR
We analyze the Periodic Decentralized Stochastic Gradient Descent (PD-SGD), a straightforward combination of federated averaging and decentralized SGD. Expand
  • 17
  • 4