Search Results - "Kale, Satyen"

Logarithmic regret algorithms for online convex optimization by HAZAN, Elad, AGARWAL, Amit, KALE, Satyen

“…Issue Title: Special Issue on COLT 2006; Guest Editors: Avrim Blum, Gabor Lugosi and Hans Ulrich Simon In an online convex optimization problem a…”
Get full text

Extracting certainty from uncertainty: regret bounded by variation in costs by Hazan, Elad, Kale, Satyen

“…Prediction from expert advice is a fundamental problem in machine learning. A major pillar of the field is the existence of learning algorithms whose average…”
Get full text

AN ONLINE PORTFOLIO SELECTION ALGORITHM WITH REGRET LOGARITHMIC IN PRICE VARIATION by Hazan, Elad, Kale, Satyen

“…We present a novel efficient algorithm for portfolio selection which theoretically attains two desirable properties: Worst‐case guarantee: the algorithm is…”
Get full text

Learning rotations with little regret by Hazan, Elad, Kale, Satyen, Warmuth, Manfred K.

“…We describe online algorithms for learning a rotation from pairs of unit vectors in R n . We show that the expected regret of our online algorithm compared to…”
Get full text

Extracting certainty from uncertainty: regret bounded byvariation incosts by Hazan, Elad, Kale, Satyen

“…Prediction from expert advice is a fundamental problem in machine learning. A major pillar of the field is the existence of learning algorithms whose average…”
Get full text

Extracting certainty from uncertainty: regret bounded by variation in costs: Special Issue on Learning Theory by HAZAN, Elad, KALE, Satyen

Get full text

A variation on SVD based image compression by Ranade, Abhiram, Mahabalarao, Srikanth S., Kale, Satyen

“…We present a variation to the well studied SVD based image compression technique. Our variation can be viewed as a preprocessing step in which the input image…”
Get full text

Stacking as Accelerated Gradient Descent by Agarwal, Naman, Awasthi, Pranjal, Kale, Satyen, Zhao, Eric

“…Stacking, a heuristic technique for training deep residual networks by progressively increasing the number of layers and initializing new layers by copying…”
Get full text

Improved Differentially Private and Lazy Online Convex Optimization by Agarwal, Naman, Kale, Satyen, Singh, Karan, Thakurta, Abhradeep Guha

“…We study the task of $(\epsilon, \delta)$-differentially private online convex optimization (OCO). In the online setting, the release of each distinct decision…”
Get full text

Pushing the Efficiency-Regret Pareto Frontier for Online Learning of Portfolios and Quantum States by Zimmert, Julian, Agarwal, Naman, Kale, Satyen

“…We revisit the classical online portfolio selection problem. It is widely assumed that a trade-off between computational complexity and regret is unavoidable,…”
Get full text

Almost Tight Bounds for Differentially Private Densest Subgraph by Dinitz, Michael, Kale, Satyen, Lattanzi, Silvio, Vassilvitskii, Sergei

“…We study the Densest Subgraph (DSG) problem under the additional constraint of differential privacy. DSG is a fundamental theoretical question which plays a…”
Get full text

Efficient Methods for Online Multiclass Logistic Regression by Agarwal, Naman, Kale, Satyen, Zimmert, Julian

“…Multiclass logistic regression is a fundamental task in machine learning with applications in classification and boosting. Previous work (Foster et al., 2018)…”
Get full text

SGD: The Role of Implicit Regularization, Batch-size and Multiple-epochs by Kale, Satyen, Sekhari, Ayush, Sridharan, Karthik

“…Multi-epoch, small-batch, Stochastic Gradient Descent (SGD) has been the method of choice for learning with large over-parameterized models. A popular theory…”
Get full text

A Multiclass Boosting Framework for Achieving Fast and Provable Adversarial Robustness by Abernethy, Jacob, Awasthi, Pranjal, Kale, Satyen

“…Alongside the well-publicized accomplishments of deep neural networks there has emerged an apparent bug in their success on tasks such as object recognition:…”
Get full text

Beyond Uniform Lipschitz Condition in Differentially Private Optimization by Das, Rudrajit, Kale, Satyen, Xu, Zheng, Zhang, Tong, Sanghavi, Sujay

“…Most prior results on differentially private stochastic gradient descent (DP-SGD) are derived under the simplistic assumption of uniform Lipschitzness, i.e.,…”
Get full text

Efficient Stagewise Pretraining via Progressive Subnetworks by Panigrahi, Abhishek, Saunshi, Nikunj, Lyu, Kaifeng, Miryoosefi, Sobhan, Reddi, Sashank, Kale, Satyen, Kumar, Sanjiv

“…Recent developments in large language models have sparked interest in efficient pretraining methods. Stagewise training approaches to improve efficiency, like…”
Get full text

Agnostic Learnability of Halfspaces via Logistic Loss by Ji, Ziwei, Ahn, Kwangjun, Awasthi, Pranjal, Kale, Satyen, Karp, Stefani

“…We investigate approximation guarantees provided by logistic regression for the fundamental problem of agnostic learning of homogeneous halfspaces. Previously,…”
Get full text

A Deep Conditioning Treatment of Neural Networks by Agarwal, Naman, Awasthi, Pranjal, Kale, Satyen

“…We study the role of depth in training randomly initialized overparameterized neural networks. We give a general result showing that depth improves…”
Get full text

Multiarmed Bandits With Limited Expert Advice by Kale, Satyen

“…We solve the COLT 2013 open problem of \citet{SCB} on minimizing regret in the setting of advice-efficient multiarmed bandits with expert advice. We give an…”
Get full text

Asynchronous Local-SGD Training for Language Modeling by Liu, Bo, Chhaparia, Rachita, Douillard, Arthur, Kale, Satyen, Rusu, Andrei A, Shen, Jiajun, Szlam, Arthur, Ranzato, Marc'Aurelio

“…Local stochastic gradient descent (Local-SGD), also referred to as federated averaging, is an approach to distributed optimization where each device performs…”
Get full text