72 citations · 542 across the 65 of their papers we have counts for
6 papers · 1 filter
Minimum weight norm models do not always generalize well for over-parameterized problems
Vatsal Shah, Anastasios Kyrillidis, Sujay Sanghavi
This work is substituted by the paper in arXiv:2011.14066. Stochastic gradient descent is the de facto algorithm for training deep neural networks (DNNs). Despite its popularity, i…
Validating and Certifying Stabilizer States
Amir Kalev, Anastasios Kyrillidis, Norbert M. Linke
We propose a measurement scheme that validates the preparation of an -qubit stabilizer state. The scheme involves a measurement of Pauli observables, a priori determined fro…
Implicit regularization and solution uniqueness in over-parameterized matrix sensing
Kelly Geyer, Anastasios Kyrillidis, Amir Kalev
We consider whether algorithmic choices in over-parameterized linear matrix factorization introduce implicit regularization. We focus on noiseless matrix sensing over rank- posi…
Provably convergent acceleration in factored gradient descent with applications in matrix sensing
Tayo Ajayi, David Mildebrath, Anastasios Kyrillidis +3
We present theoretical results on the convergence of \emph{non-convex} accelerated gradient descent in matrix factorization models with -norm loss. The purpose of this work…
Simple and practical algorithms for -norm low-rank approximation
Anastasios Kyrillidis
We propose practical algorithms for entrywise -norm low-rank approximation, for or . The proposed framework, which is non-convex and gradient-based, is…
Approximate Newton-based statistical inference using only stochastic gradients
Tianyang Li, Anastasios Kyrillidis, Liu Liu +1
We present a novel statistical inference framework for convex empirical risk minimization, using approximate stochastic Newton steps. The proposed algorithm is based on the notion…