• Publications
  • Influence
XORing Elephants: Novel Erasure Codes for Big Data
TLDR
We present a novel family of erasure codes that are efficiently repairable and offer higher reliability compared to Reed-Solomon codes. Expand
  • 625
  • 84
  • PDF
Speeding Up Distributed Machine Learning Using Codes
TLDR
Codes are widely used in many engineering applications to offer <italic>robustness</italic>. Expand
  • 333
  • 58
  • PDF
Perturbed Iterate Analysis for Asynchronous Stochastic Optimization
TLDR
We introduce and analyze stochastic optimization methods where the input to each gradient update is perturbed by bounded noise, that are simpler than earlier analyses, remove many assumptions of previous models. Expand
  • 143
  • 29
  • PDF
Locality and Availability in Distributed Storage
TLDR
This paper studies the problem of information symbol availability in codes: we refer to a systematic code as code with (r, t)-availability if every information (systematic) symbol can be reconstructed from t disjoint groups of other code symbols. Expand
  • 178
  • 26
  • PDF
Interference Alignment as a Rank Constrained Rank Minimization
TLDR
We show that the maximization of the sum degrees-of-freedom for the static flat-fading multiple-input multiple-output (MIMO) interference channel (IC) is equivalent to a rank constrained rank minimization problem when the signal subspaces span all available dimensions. Expand
  • 141
  • 26
  • PDF
Locally Repairable Codes
TLDR
In this paper, we explore the repair metric of locality, which corresponds to the number of disk accesses required during a single node repair. Expand
  • 264
  • 25
  • PDF
Speeding up distributed machine learning using codes
TLDR
We view distributed machine learning algorithms through a coding-theoretic lens, and show how codes can equip them with robustness against this system noise. Expand
  • 137
  • 25
  • PDF
Locally Repairable Codes
TLDR
We explore the repair metric of locality, which corresponds to the number of disk accesses required during a single node repair. Expand
  • 172
  • 17
Repair optimal erasure codes through hadamard designs
TLDR
We use Hadamard matrices to construct the first explicit 2-parity MDS storage code with optimal repair properties for all single node failures, including the parity nodes. Expand
  • 154
  • 14
  • PDF
DRACO: Byzantine-resilient Distributed Training via Redundant Gradients
TLDR
We present DRACO, a general distributed training framework that is robust against adversarial nodes and worst-case compute errors. Expand
  • 99
  • 13
  • PDF
...
1
2
3
4
5
...