P
Praneeth Netrapalli
Researcher at Microsoft
Publications - 117
Citations - 6792
Praneeth Netrapalli is an academic researcher from Microsoft. The author has contributed to research in topics: Stochastic gradient descent & Gradient descent. The author has an hindex of 38, co-authored 117 publications receiving 5387 citations. Previous affiliations of Praneeth Netrapalli include University of Texas at Austin & Google.
Papers
More filters
Proceedings Article
Online Non-Convex Learning: Following the Perturbed Leader is Optimal
TL;DR: This work shows that the classical Follow the Perturbed Leader (FTPL) algorithm achieves optimal regret rate of $O(T^{-1/2})$ in this setting, which improves upon the previous best-known regret rate for FTPL.
Posted Content
Parallelizing Stochastic Gradient Descent for Least Squares Regression: mini-batching, averaging, and model misspecification.
TL;DR: In this paper, the benefits of averaging schemes widely used in conjunction with stochastic gradient descent (SGD) were analyzed for the least square regression problem and the authors provided non-asymptotic excess risk bounds for these schemes.
Journal ArticleDOI
P-SIF: Document Embeddings Using Partition Averaging
TL;DR: P-SIF, a partitioned word averaging model to represent long documents that retains the simplicity of simple weighted word averaging while taking a document's topical structure into account and concatenates them all to represent the overall document.
Proceedings ArticleDOI
Learning Markov graphs up to edit distance
TL;DR: Lower bounds on the number of samples required for any algorithm to learn the Markov graph structure of a probability distribution, up to edit distance are provided and show that substantial gains in sample complexity may not be possible without paying a significant price in edit distance error.
Proceedings Article
Faster eigenvector computation via shift-and-invert preconditioning
Dan Garber,Elad Hazan,Chi Jin,Sham M. Kakade,Cameron Musco,Praneeth Netrapalli,Aaron Sidford +6 more
TL;DR: In this paper, the authors give a faster algorithm for estimating the top eigenvector of an explicit matrix A ∈ Rn×d in time O ([nnz(A) + d sr(A)/gap2] log 1/e ) where nnz(A), is the number of nonzeros in A, s is the stable rank, and gap is the relative eigengap.