Search Results - "Frostig, Roy"

Parente2: a fast and accurate method for detecting identity by descent by Rodriguez, Jesse M, Bercovici, Sivan, Huang, Lin, Frostig, Roy, Batzoglou, Serafim

“…Identity-by-descent (IBD) inference is the problem of establishing a genetic connection between two individuals through a genomic segment that is inherited by…”
Get full text

Lightweight Statistical Learning: Accelerating and Avoiding Empirical Risk Minimization by Frostig, Roy

“…In statistical machine learning, the goal is to train a model that, once deployed in the world, continues to predict accurately on fresh data. A unifying…”
Get full text

Learning from many trajectories by Tu, Stephen, Frostig, Roy, Soltanolkotabi, Mahdi

“…We initiate a study of supervised learning from many independent sequences ("trajectories") of non-independent covariates, reflecting tasks in sequence…”
Get full text

You Only Linearize Once: Tangents Transpose to Gradients by Radul, Alexey, Paszke, Adam, Frostig, Roy, Johnson, Matthew, Maclaurin, Dougal

“…Automatic differentiation (AD) is conventionally understood as a family of distinct algorithms, rooted in two "modes" -- forward and reverse -- which are…”
Get full text

The advantages of multiple classes for reducing overfitting from test set reuse by Feldman, Vitaly, Frostig, Roy, Hardt, Moritz

“…Excessive reuse of holdout data can lead to overfitting. However, there is little concrete evidence of significant overfitting due to holdout reuse in popular…”
Get full text

Decomposing reverse-mode automatic differentiation by Frostig, Roy, Johnson, Matthew J, Maclaurin, Dougal, Paszke, Adam, Radul, Alexey

“…We decompose reverse-mode automatic differentiation into (forward-mode) linearization followed by transposition. Doing so isolates the essential difference…”
Get full text

Efficient and Modular Implicit Differentiation by Blondel, Mathieu, Berthet, Quentin, Cuturi, Marco, Frostig, Roy, Hoyer, Stephan, Llinares-López, Felipe, Pedregosa, Fabian, Vert, Jean-Philippe

“…Automatic differentiation (autodiff) has revolutionized machine learning. It allows to express complex computations by composing elementary ones in creative…”
Get full text

Learning Model Predictive Controllers with Real-Time Attention for Real-World Navigation by Xiao, Xuesu, Zhang, Tingnan, Choromanski, Krzysztof, Lee, Edward, Francis, Anthony, Varley, Jake, Tu, Stephen, Singh, Sumeet, Xu, Peng, Xia, Fei, Persson, Sven Mikael, Kalashnikov, Dmitry, Takayama, Leila, Frostig, Roy, Tan, Jie, Parada, Carolina, Sindhwani, Vikas

“…Despite decades of research, existing navigation systems still face real-world challenges when deployed in the wild, e.g., in cluttered home environments or in…”
Get full text

A sub-constant improvement in approximating the positive semidefinite Grothendieck problem by Frostig, Roy, Wang, Sida I

“…Semidefinite relaxations are a powerful tool for approximately solving combinatorial optimization problems such as MAX-CUT and the Grothendieck problem. By…”
Get full text

Toward Deeper Understanding of Neural Networks: The Power of Initialization and a Dual View on Expressivity by Daniely, Amit, Frostig, Roy, Singer, Yoram

“…We develop a general duality between neural networks and compositional kernels, striving towards a better understanding of deep learning. We show that initial…”
Get full text

Random Features for Compositional Kernels by Daniely, Amit, Frostig, Roy, Gupta, Vineet, Singer, Yoram

“…We describe and analyze a simple random feature scheme (RFS) from prescribed compositional kernels. The compositional kernels we use are inspired by the…”
Get full text

Measuring the Effects of Data Parallelism on Neural Network Training by Shallue, Christopher J, Lee, Jaehoon, Antognini, Joseph, Sohl-Dickstein, Jascha, Frostig, Roy, Dahl, George E

“…Journal of Machine Learning Research 20 (2019) 1-49 Recent hardware developments have dramatically increased the scale of data parallelism available for neural…”
Get full text

Estimation from Indirect Supervision with Linear Moments by Raghunathan, Aditi, Frostig, Roy, Duchi, John, Liang, Percy

“…In structured prediction problems where we have indirect supervision of the output, maximum marginal likelihood faces two computational obstacles:…”
Get full text

Principal Component Projection Without Principal Component Analysis by Frostig, Roy, Musco, Cameron, Musco, Christopher, Sidford, Aaron

“…We show how to efficiently project a vector onto the top principal components of a matrix, without explicitly computing these components. Specifically, we…”
Get full text

Un-regularizing: approximate proximal point and faster stochastic algorithms for empirical risk minimization by Frostig, Roy, Ge, Rong, Kakade, Sham M, Sidford, Aaron

“…We develop a family of accelerated stochastic algorithms that minimize sums of convex functions. Our algorithms improve upon the fastest running time for…”
Get full text

Competing with the Empirical Risk Minimizer in a Single Pass by Frostig, Roy, Ge, Rong, Kakade, Sham M, Sidford, Aaron

“…In many estimation problems, e.g. linear and logistic regression, we wish to minimize an unknown objective given only unbiased samples of the objective…”
Get full text

Relaxations for inference in restricted Boltzmann machines by Wang, Sida I, Frostig, Roy, Liang, Percy, Manning, Christopher D

“…We propose a relaxation-based approximate inference algorithm that samples near-MAP configurations of a binary pairwise Markov random field. We experiment on…”
Get full text