Roy Frostig
Roy Frostig
Research scientist, Google Research, Brain team
Verified email at cs.stanford.edu - Homepage
TitleCited byYear
Semantic Parsing on Freebase from Question-Answer Pairs
J Berant, A Chou, R Frostig, P Liang
Empirical methods in natural language processing (EMNLP), 2013
7992013
Un-regularizing: approximate proximal point and faster stochastic algorithms for empirical risk minimization
R Frostig, R Ge, SM Kakade, A Sidford
Proceedings of The 32nd International Conference on Machine Learning, 2540-2548, 2015
1042015
Toward deeper understanding of neural networks: The power of initialization and a dual view on expressivity
A Daniely, R Frostig, Y Singer
Advances In Neural Information Processing Systems, 2253-2261, 2016
1032016
Competing with the empirical risk minimizer in a single pass
R Frostig, R Ge, SM Kakade, A Sidford
Conference on learning theory, 728-763, 2015
672015
Measuring the Effects of Data Parallelism on Neural Network Training
CJ Shallue, J Lee, J Antognini, J Sohl-Dickstein, R Frostig, GE Dahl
Journal of Machine Learning Research 20 (112), 1-49, 2019
442019
Principal component projection without principal component analysis
R Frostig, C Musco, C Musco, A Sidford
International Conference on Machine Learning, 2349-2357, 2016
182016
Simple MAP Inference via Low-Rank Relaxations
R Frostig, S Wang, PS Liang, CD Manning
Advances in Neural Information Processing Systems, 3077-3085, 2014
142014
The advantages of multiple classes for reducing overfitting from test set reuse
V Feldman, R Frostig, M Hardt
International Conference on Machine Learning, 1892-1900, 2019
62019
Random Features for Compositional Kernels
A Daniely, R Frostig, V Gupta, Y Singer
arXiv preprint arXiv:1703.07872, 2017
62017
Relaxations for inference in restricted Boltzmann machines
SI Wang, R Frostig, P Liang, CD Manning
arXiv preprint arXiv:1312.6205, 2013
52013
Compiling machine learning programs via high-level tracing
R Frostig, MJ Johnson, C Leary
SysML, 2018
32018
Estimation from Indirect Supervision with Linear Moments
A Raghunathan, R Frostig, J Duchi, P Liang
arXiv preprint arXiv:1608.03100, 2016
32016
Open Problem: How fast can a multiclass test set be overfit?
V Feldman, R Frostig, M Hardt
Conference on Learning Theory, 3185-3189, 2019
12019
A sub-constant improvement in approximating the positive semidefinite Grothendieck problem
R Frostig, SI Wang
arXiv preprint arXiv:1408.2270, 2014
2014
The system can't perform the operation now. Try again later.
Articles 1–14