Get our free extension to see links to code for papers anywhere online!

Chrome logo Add to Chrome

Firefox logo Add to Firefox

Picture for Mikhail Belkin

Local Quadratic Convergence of Stochastic Gradient Descent with Adaptive Step Size


Dec 30, 2021
Adityanarayanan Radhakrishnan, Mikhail Belkin, Caroline Uhler

* ICML 2021 Workshop on Beyond first-order methods in ML systems 

  Access Paper or Ask Questions

Simple, Fast, and Flexible Framework for Matrix Completion with Infinite Width Neural Networks


Jul 31, 2021
Adityanarayanan Radhakrishnan, George Stefanakis, Mikhail Belkin, Caroline Uhler


  Access Paper or Ask Questions

Fit without fear: remarkable mathematical phenomena of deep learning through the prism of interpolation


May 29, 2021
Mikhail Belkin

* A version of this paper will appear in Acta Numerica 

  Access Paper or Ask Questions

Risk Bounds for Over-parameterized Maximum Margin Classification on Sub-Gaussian Mixtures


Apr 28, 2021
Yuan Cao, Quanquan Gu, Mikhail Belkin

* 35 pages, 3 figures 

  Access Paper or Ask Questions

On the linearity of large non-linear models: when and why the tangent kernel is constant


Oct 02, 2020
Chaoyue Liu, Libin Zhu, Mikhail Belkin


  Access Paper or Ask Questions

Linear Convergence and Implicit Regularization of Generalized Mirror Descent with Time-Dependent Mirrors


Sep 18, 2020
Adityanarayanan Radhakrishnan, Mikhail Belkin, Caroline Uhler


  Access Paper or Ask Questions

Multiple Descent: Design Your Own Generalization Curve


Aug 12, 2020
Lin Chen, Yifei Min, Mikhail Belkin, Amin Karbasi

* Fixed minor typos 

  Access Paper or Ask Questions

Evaluation of Neural Architectures Trained with Square Loss vs Cross-Entropy in Classification Tasks


Jun 12, 2020
Like Hui, Mikhail Belkin

* 13 pages, 1 figure with 3 subfigures, 15 tables 

  Access Paper or Ask Questions

Classification vs regression in overparameterized regimes: Does the loss function matter?


May 16, 2020
Vidya Muthukumar, Adhyyan Narang, Vignesh Subramanian, Mikhail Belkin, Daniel Hsu, Anant Sahai


  Access Paper or Ask Questions

Toward a theory of optimization for over-parameterized systems of non-linear equations: the lessons of deep learning


Feb 29, 2020
Chaoyue Liu, Libin Zhu, Mikhail Belkin


  Access Paper or Ask Questions

Overparameterized Neural Networks Can Implement Associative Memory


Sep 26, 2019
Adityanarayanan Radhakrishnan, Mikhail Belkin, Caroline Uhler


  Access Paper or Ask Questions

Two models of double descent for weak features


Mar 18, 2019
Mikhail Belkin, Daniel Hsu, Ji Xu


  Access Paper or Ask Questions

Reconciling modern machine learning and the bias-variance trade-off


Dec 28, 2018
Mikhail Belkin, Daniel Hsu, Siyuan Ma, Soumik Mandal


  Access Paper or Ask Questions

On exponential convergence of SGD in non-convex over-parametrized learning


Nov 06, 2018
Raef Bassily, Mikhail Belkin, Siyuan Ma


  Access Paper or Ask Questions

Kernel Machines Beat Deep Neural Networks on Mask-based Single-channel Speech Enhancement


Nov 06, 2018
Like Hui, Siyuan Ma, Mikhail Belkin


  Access Paper or Ask Questions

MaSS: an Accelerated Stochastic Method for Over-parametrized Learning


Nov 05, 2018
Chaoyue Liu, Mikhail Belkin

* This is version 2. Find typos in version 1 .Make the corrections 

  Access Paper or Ask Questions

Overfitting or perfect fitting? Risk bounds for classification and regression rules that interpolate


Oct 26, 2018
Mikhail Belkin, Daniel Hsu, Partha Mitra


  Access Paper or Ask Questions

Learning kernels that adapt to GPU


Oct 19, 2018
Siyuan Ma, Mikhail Belkin


  Access Paper or Ask Questions

Downsampling leads to Image Memorization in Convolutional Autoencoders


Oct 16, 2018
Adityanarayanan Radhakrishnan, Mikhail Belkin, Caroline Uhler


  Access Paper or Ask Questions

Approximation beats concentration? An approximation view on inference with smooth radial kernels


Aug 02, 2018
Mikhail Belkin

* Conference on Computational Learning Theory (COLT) 2018 

  Access Paper or Ask Questions

Does data interpolation contradict statistical optimality?


Jun 25, 2018
Mikhail Belkin, Alexander Rakhlin, Alexandre B. Tsybakov


  Access Paper or Ask Questions

To understand deep learning we need to understand kernel learning


Jun 14, 2018
Mikhail Belkin, Siyuan Ma, Soumik Mandal


  Access Paper or Ask Questions

The Power of Interpolation: Understanding the Effectiveness of SGD in Modern Over-parametrized Learning


Jun 14, 2018
Siyuan Ma, Raef Bassily, Mikhail Belkin


  Access Paper or Ask Questions

Parametrized Accelerated Methods Free of Condition Number


Feb 28, 2018
Chaoyue Liu, Mikhail Belkin

* 23 pages, 3 figures 

  Access Paper or Ask Questions

Eigenvectors of Orthogonally Decomposable Functions


Feb 23, 2018
Mikhail Belkin, Luis Rademacher, James Voss

* 69 pages 

  Access Paper or Ask Questions

Fast Interactive Image Retrieval using large-scale unlabeled data


Feb 12, 2018
Akshay Mehra, Jihun Hamm, Mikhail Belkin

* 15 Pages, Submitted to KDD 2018 

  Access Paper or Ask Questions

Unperturbed: spectral analysis beyond Davis-Kahan


Jun 20, 2017
Justin Eldridge, Mikhail Belkin, Yusu Wang


  Access Paper or Ask Questions

Diving into the shallows: a computational perspective on large-scale shallow learning


Jun 17, 2017
Siyuan Ma, Mikhail Belkin


  Access Paper or Ask Questions

Graphons, mergeons, and so on!


May 22, 2017
Justin Eldridge, Mikhail Belkin, Yusu Wang


  Access Paper or Ask Questions