Timezone: »
While stochastic gradient descent (SGD) is still the de facto algorithm in deep learning, adaptive methods like Clipped SGD/Adam have been observed to outperform SGD across important tasks, such as attention models. The settings under which SGD performs poorly in comparison to adaptive methods are not well understood yet. In this paper, we provide empirical and theoretical evidence that a heavy-tailed distribution of the noise in stochastic gradients is one cause of SGD's poor performance. We provide the first tight upper and lower convergence bounds for adaptive gradient methods under heavy-tailed noise. Further, we demonstrate how gradient clipping plays a key role in addressing heavy-tailed gradient noise. Subsequently, we show how clipping can be applied in practice by developing an adaptive coordinate-wise clipping algorithm (ACClip) and demonstrate its superior performance on BERT pretraining and finetuning tasks.
Author Information
Jingzhao Zhang (MIT)
Sai Praneeth Karimireddy (EPFL)
I am a second year PhD student working in convex and non-convex optimization with Prof. Martin Jaggi. My focus is on designing faster and more scalable optimization algorithms for machine learning. Some of my preliminary results and problems I am currently working on- 1. Robust accelerated algorithms - Nesterov acceleration modified to be robust to noise. 2. Faster algorithms which take second order information about the function into account. 3. A $O(1/t^2)$ rate *affine invariant* algorithm for constrained optimization. 4. Frank-Wolfe algorithm for non-smooth functions using 'noisy-smoothing'
Andreas Veit (Google)
Seungyeon Kim (Google Research)
Sashank Reddi (Google)
Sanjiv Kumar (Google Research)
Suvrit Sra (MIT)
Suvrit Sra is a faculty member within the EECS department at MIT, where he is also a core faculty member of IDSS, LIDS, MIT-ML Group, as well as the statistics and data science center. His research spans topics in optimization, matrix theory, differential geometry, and probability theory, which he connects with machine learning --- a key focus of his research is on the theme "Optimization for Machine Learning” (http://opt-ml.org)
More from the Same Authors
-
2021 : An Empirical Study of Pre-trained Models on Out-of-distribution Generalization »
Yaodong Yu · Heinrich Jiang · Dara Bahri · Hossein Mobahi · Seungyeon Kim · Ankit Rawat · Andreas Veit · Yi Ma -
2022 : Differentially Private Adaptive Optimization with Delayed Preconditioners »
Tian Li · Manzil Zaheer · Ken Liu · Sashank Reddi · H. Brendan McMahan · Virginia Smith -
2022 : Differentially Private Adaptive Optimization with Delayed Preconditioners »
Tian Li · Manzil Zaheer · Ken Liu · Sashank Reddi · H. Brendan McMahan · Virginia Smith -
2022 : Effect of mixup Training on Representation Learning »
Arslan Chaudhry · Aditya Menon · Andreas Veit · Sadeep Jayasumana · Srikumar Ramalingam · Sanjiv Kumar -
2023 Poster: The Curious Role of Normalization in Sharpness-Aware Minimization »
Yan Dai · Kwangjun Ahn · Suvrit Sra -
2023 Poster: SOAR: Improved Quantization for Nearest Neighbor Search »
Philip Sun · David Simcha · Dave Dopson · Ruiqi Guo · Sanjiv Kumar -
2023 Poster: Transformers learn to implement preconditioned gradient descent for in-context learning »
Kwangjun Ahn · Xiang Cheng · Hadi Daneshmand · Suvrit Sra -
2023 Poster: ResMem: Learn what you can and memorize the rest »
Zitong Yang · MICHAL LUKASIK · Vaishnavh Nagarajan · Zonglin Li · Ankit Rawat · Manzil Zaheer · Aditya Menon · Sanjiv Kumar -
2023 Poster: On student-teacher deviations in distillation: does it pay to disobey? »
Vaishnavh Nagarajan · Aditya Menon · Srinadh Bhojanapalli · Hossein Mobahi · Sanjiv Kumar -
2023 Poster: When Does Confidence-Based Cascade Deferral Suffice? »
Wittawat Jitkrittum · Neha Gupta · Aditya Menon · Harikrishna Narasimhan · Ankit Rawat · Sanjiv Kumar -
2023 Poster: What is the Inductive Bias of Flatness Regularization? A Study of Deep Matrix Factorization Models »
Khashayar Gatmiry · Zhiyuan Li · Tengyu Ma · Sashank Reddi · Stefanie Jegelka · Ching-Yao Chuang -
2022 Poster: TPU-KNN: K Nearest Neighbor Search at Peak FLOP/s »
Felix Chern · Blake Hechtman · Andy Davis · Ruiqi Guo · David Majnemer · Sanjiv Kumar -
2022 Poster: CCCP is Frank-Wolfe in disguise »
Alp Yurtsever · Suvrit Sra -
2022 Poster: Decoupled Context Processing for Context Augmented Language Modeling »
Zonglin Li · Ruiqi Guo · Sanjiv Kumar -
2022 Poster: Efficient Sampling on Riemannian Manifolds via Langevin MCMC »
Xiang Cheng · Jingzhao Zhang · Suvrit Sra -
2022 Poster: Post-hoc estimators for learning to defer to an expert »
Harikrishna Narasimhan · Wittawat Jitkrittum · Aditya Menon · Ankit Rawat · Sanjiv Kumar -
2021 Poster: Batch Active Learning at Scale »
Gui Citovsky · Giulia DeSalvo · Claudio Gentile · Lazaros Karydas · Anand Rajagopalan · Afshin Rostamizadeh · Sanjiv Kumar -
2021 Poster: Can contrastive learning avoid shortcut solutions? »
Joshua Robinson · Li Sun · Ke Yu · Kayhan Batmanghelich · Stefanie Jegelka · Suvrit Sra -
2021 Poster: Breaking the centralized barrier for cross-device federated learning »
Sai Praneeth Karimireddy · Martin Jaggi · Satyen Kale · Mehryar Mohri · Sashank Reddi · Sebastian Stich · Ananda Theertha Suresh -
2021 Poster: Efficient Training of Retrieval Models using Negative Cache »
Erik Lindgren · Sashank Reddi · Ruiqi Guo · Sanjiv Kumar -
2021 Poster: Complexity Lower Bounds for Nonconvex-Strongly-Concave Min-Max Optimization »
Haochuan Li · Yi Tian · Jingzhao Zhang · Ali Jadbabaie -
2021 Poster: Three Operator Splitting with Subgradients, Stochastic Gradients, and Adaptive Learning Rates »
Alp Yurtsever · Alex Gu · Suvrit Sra -
2021 Poster: Fast Federated Learning in the Presence of Arbitrary Device Unavailability »
Xinran Gu · Kaixuan Huang · Jingzhao Zhang · Longbo Huang -
2021 Poster: RelaySum for Decentralized Deep Learning on Heterogeneous Data »
Thijs Vogels · Lie He · Anastasiia Koloskova · Sai Praneeth Karimireddy · Tao Lin · Sebastian Stich · Martin Jaggi -
2020 : Invited speaker: SGD without replacement: optimal rate analysis and more, Suvrit Sra »
Suvrit Sra -
2020 Poster: SGD with shuffling: optimal rates without component convexity and large epoch requirements »
Kwangjun Ahn · Chulhee Yun · Suvrit Sra -
2020 Spotlight: SGD with shuffling: optimal rates without component convexity and large epoch requirements »
Kwangjun Ahn · Chulhee Yun · Suvrit Sra -
2020 Poster: Practical Low-Rank Communication Compression in Decentralized Deep Learning »
Thijs Vogels · Sai Praneeth Karimireddy · Martin Jaggi -
2020 Poster: Multi-Stage Influence Function »
Hongge Chen · Si Si · Yang Li · Ciprian Chelba · Sanjiv Kumar · Duane Boning · Cho-Jui Hsieh -
2020 Poster: O(n) Connections are Expressive Enough: Universal Approximability of Sparse Transformers »
Chulhee Yun · Yin-Wen Chang · Srinadh Bhojanapalli · Ankit Singh Rawat · Sashank Reddi · Sanjiv Kumar -
2020 Poster: Robust large-margin learning in hyperbolic space »
Melanie Weber · Manzil Zaheer · Ankit Singh Rawat · Aditya Menon · Sanjiv Kumar -
2020 Poster: Towards Minimax Optimal Reinforcement Learning in Factored Markov Decision Processes »
Yi Tian · Jian Qian · Suvrit Sra -
2020 Poster: Learning discrete distributions: user vs item-level privacy »
Yuhan Liu · Ananda Theertha Suresh · Felix Xinnan Yu · Sanjiv Kumar · Michael D Riley -
2020 Spotlight: Towards Minimax Optimal Reinforcement Learning in Factored Markov Decision Processes »
Yi Tian · Jian Qian · Suvrit Sra -
2019 : Poster Session »
Eduard Gorbunov · Alexandre d'Aspremont · Lingxiao Wang · Liwei Wang · Boris Ginsburg · Alessio Quaglino · Camille Castera · Saurabh Adya · Diego Granziol · Rudrajit Das · Raghu Bollapragada · Fabian Pedregosa · Martin Takac · Majid Jahani · Sai Praneeth Karimireddy · Hilal Asi · Balint Daroczy · Leonard Adolphs · Aditya Rawal · Nicolas Brandt · Minhan Li · Giuseppe Ughi · Orlando Romero · Ivan Skorokhodov · Damien Scieur · Kiwook Bae · Konstantin Mishchenko · Rohan Anil · Vatsal Sharan · Aditya Balu · Chao Chen · Zhewei Yao · Tolga Ergen · Paul Grigas · Chris Junchi Li · Jimmy Ba · Stephen J Roberts · Sharan Vaswani · Armin Eftekhari · Chhavi Sharma -
2019 Poster: Breaking the Glass Ceiling for Embedding-Based Classifiers for Large Output Spaces »
Chuan Guo · Ali Mousavi · Xiang Wu · Daniel Holtmann-Rice · Satyen Kale · Sashank Reddi · Sanjiv Kumar -
2019 Poster: PowerSGD: Practical Low-Rank Gradient Compression for Distributed Optimization »
Thijs Vogels · Sai Praneeth Karimireddy · Martin Jaggi -
2019 Poster: Flexible Modeling of Diversity with Strongly Log-Concave Distributions »
Joshua Robinson · Suvrit Sra · Stefanie Jegelka -
2019 Poster: Are deep ResNets provably better than linear predictors? »
Chulhee Yun · Suvrit Sra · Ali Jadbabaie -
2019 Poster: Multilabel reductions: what is my loss optimising? »
Aditya Menon · Ankit Singh Rawat · Sashank Reddi · Sanjiv Kumar -
2019 Poster: Small ReLU networks are powerful memorizers: a tight analysis of memorization capacity »
Chulhee Yun · Suvrit Sra · Ali Jadbabaie -
2019 Spotlight: Multilabel reductions: what is my loss optimising? »
Aditya Menon · Ankit Singh Rawat · Sashank Reddi · Sanjiv Kumar -
2019 Spotlight: Small ReLU networks are powerful memorizers: a tight analysis of memorization capacity »
Chulhee Yun · Suvrit Sra · Ali Jadbabaie -
2019 Poster: Sampled Softmax with Random Fourier Features »
Ankit Singh Rawat · Jiecao Chen · Felix Xinnan Yu · Ananda Theertha Suresh · Sanjiv Kumar -
2018 Poster: Direct Runge-Kutta Discretization Achieves Acceleration »
Jingzhao Zhang · Aryan Mokhtari · Suvrit Sra · Ali Jadbabaie -
2018 Spotlight: Direct Runge-Kutta Discretization Achieves Acceleration »
Jingzhao Zhang · Aryan Mokhtari · Suvrit Sra · Ali Jadbabaie -
2018 Poster: Adaptive Methods for Nonconvex Optimization »
Manzil Zaheer · Sashank Reddi · Devendra S Sachan · Satyen Kale · Sanjiv Kumar -
2018 Poster: Exponentiated Strongly Rayleigh Distributions »
Zelda Mariet · Suvrit Sra · Stefanie Jegelka -
2018 Poster: cpSGD: Communication-efficient and differentially-private distributed SGD »
Naman Agarwal · Ananda Theertha Suresh · Felix Xinnan Yu · Sanjiv Kumar · Brendan McMahan -
2018 Spotlight: cpSGD: Communication-efficient and differentially-private distributed SGD »
Naman Agarwal · Ananda Theertha Suresh · Felix Xinnan Yu · Sanjiv Kumar · Brendan McMahan -
2018 Tutorial: Negative Dependence, Stable Polynomials, and All That »
Suvrit Sra · Stefanie Jegelka -
2017 : Now Playing: Continuous low-power music recognition »
Marvin Ritter · Ruiqi Guo · Sanjiv Kumar · Julian J Odell · Mihajlo Velimirović · Dominik Roblek · James Lyon -
2017 Workshop: OPT 2017: Optimization for Machine Learning »
Suvrit Sra · Sashank J. Reddi · Alekh Agarwal · Benjamin Recht -
2017 Poster: Elementary Symmetric Polynomials for Optimal Experimental Design »
Zelda Mariet · Suvrit Sra -
2017 Poster: Multiscale Quantization for Fast Similarity Search »
Xiang Wu · Ruiqi Guo · Ananda Theertha Suresh · Sanjiv Kumar · Daniel Holtmann-Rice · David Simcha · Felix Yu -
2017 Poster: Polynomial time algorithms for dual volume sampling »
Chengtao Li · Stefanie Jegelka · Suvrit Sra -
2016 Workshop: OPT 2016: Optimization for Machine Learning »
Suvrit Sra · Francis Bach · Sashank J. Reddi · Niao He -
2016 : Taming non-convexity via geometry »
Suvrit Sra -
2016 Poster: Fast Mixing Markov Chains for Strongly Rayleigh Measures, DPPs, and Constrained Sampling »
Chengtao Li · Suvrit Sra · Stefanie Jegelka -
2016 Poster: Kronecker Determinantal Point Processes »
Zelda Mariet · Suvrit Sra -
2016 Poster: Orthogonal Random Features »
Felix Xinnan Yu · Ananda Theertha Suresh · Krzysztof M Choromanski · Daniel Holtmann-Rice · Sanjiv Kumar -
2016 Oral: Orthogonal Random Features »
Felix Xinnan Yu · Ananda Theertha Suresh · Krzysztof M Choromanski · Daniel Holtmann-Rice · Sanjiv Kumar -
2016 Poster: Proximal Stochastic Methods for Nonsmooth Nonconvex Finite-Sum Optimization »
Sashank J. Reddi · Suvrit Sra · Barnabas Poczos · Alexander Smola -
2016 Poster: Riemannian SVRG: Fast Stochastic Optimization on Riemannian Manifolds »
Hongyi Zhang · Sashank J. Reddi · Suvrit Sra -
2016 Tutorial: Large-Scale Optimization: Beyond Stochastic Gradient Descent and Convexity »
Suvrit Sra · Francis Bach -
2015 Workshop: The 1st International Workshop "Feature Extraction: Modern Questions and Challenges" »
Dmitry Storcheus · Sanjiv Kumar · Afshin Rostamizadeh -
2015 Workshop: Optimization for Machine Learning (OPT2015) »
Suvrit Sra · Alekh Agarwal · Leon Bottou · Sashank J. Reddi -
2015 Poster: Spherical Random Features for Polynomial Kernels »
Jeffrey Pennington · Felix Yu · Sanjiv Kumar -
2015 Spotlight: Spherical Random Features for Polynomial Kernels »
Jeffrey Pennington · Felix Yu · Sanjiv Kumar -
2015 Poster: Structured Transforms for Small-Footprint Deep Learning »
Vikas Sindhwani · Tara Sainath · Sanjiv Kumar -
2015 Spotlight: Structured Transforms for Small-Footprint Deep Learning »
Vikas Sindhwani · Tara Sainath · Sanjiv Kumar -
2015 Poster: Matrix Manifold Optimization for Gaussian Mixtures »
Reshad Hosseini · Suvrit Sra -
2015 Poster: On Variance Reduction in Stochastic Gradient Descent and its Asynchronous Variants »
Sashank J. Reddi · Ahmed Hefny · Suvrit Sra · Barnabas Poczos · Alexander Smola -
2014 Workshop: OPT2014: Optimization for Machine Learning »
Zaid Harchaoui · Suvrit Sra · Alekh Agarwal · Martin Jaggi · Miro Dudik · Aaditya Ramdas · Jean Lasserre · Yoshua Bengio · Amir Beck -
2014 Poster: Efficient Structured Matrix Rank Minimization »
Adams Wei Yu · Wanli Ma · Yaoliang Yu · Jaime Carbonell · Suvrit Sra -
2014 Session: Oral Session 8 »
Sanjiv Kumar -
2014 Poster: Discrete Graph Hashing »
Wei Liu · Cun Mu · Sanjiv Kumar · Shih-Fu Chang -
2014 Spotlight: Discrete Graph Hashing »
Wei Liu · Cun Mu · Sanjiv Kumar · Shih-Fu Chang -
2013 Workshop: OPT2013: Optimization for Machine Learning »
Suvrit Sra · Alekh Agarwal -
2013 Poster: Geometric optimisation on positive definite matrices for elliptically contoured distributions »
Suvrit Sra · Reshad Hosseini -
2013 Poster: Reflection methods for user-friendly submodular optimization »
Stefanie Jegelka · Francis Bach · Suvrit Sra -
2012 Workshop: Optimization for Machine Learning »
Suvrit Sra · Alekh Agarwal -
2012 Poster: A new metric on the manifold of kernel matrices with application to matrix geometric means »
Suvrit Sra -
2012 Poster: Scalable nonconvex inexact proximal splitting »
Suvrit Sra -
2012 Poster: Angular Quantization based Binary Codes for Fast Similarity Search »
Yunchao Gong · Sanjiv Kumar · Vishal Verma · Svetlana Lazebnik -
2011 Workshop: Optimization for Machine Learning »
Suvrit Sra · Stephen Wright · Sebastian Nowozin -
2010 Workshop: Numerical Mathematics Challenges in Machine Learning »
Matthias Seeger · Suvrit Sra -
2010 Workshop: Optimization for Machine Learning »
Suvrit Sra · Sebastian Nowozin · Stephen Wright -
2009 Workshop: Optimization for Machine Learning »
Sebastian Nowozin · Suvrit Sra · S.V.N Vishwanthan · Stephen Wright -
2009 Poster: Ensemble Nystrom Method »
Sanjiv Kumar · Mehryar Mohri · Ameet S Talwalkar -
2008 Workshop: Optimization for Machine Learning »
Suvrit Sra · Sebastian Nowozin · Vishwanathan S V N