Sledovat
Hrayr Harutyunyan
Hrayr Harutyunyan
Research Scientist, Google Research
E-mailová adresa ověřena na: usc.edu - Domovská stránka
Název
Citace
Citace
Rok
Mixhop: Higher-order graph convolutional architectures via sparsified neighborhood mixing
S Abu-El-Haija, B Perozzi, A Kapoor, N Alipourfard, K Lerman, ...
international conference on machine learning, 21-29, 2019
10932019
Multitask learning and benchmarking with clinical time series data
H Harutyunyan, H Khachatrian, DC Kale, G Ver Steeg, A Galstyan
Scientific data 6 (1), 96, 2019
10702019
Improving generalization by controlling label-noise information in neural network weights
H Harutyunyan, K Reing, G Ver Steeg, A Galstyan
International Conference on Machine Learning, 4071-4081, 2020
642020
Information-theoretic generalization bounds for black-box learning algorithms
H Harutyunyan, M Raginsky, G Ver Steeg, A Galstyan
Advances in Neural Information Processing Systems 34, 24670-24682, 2021
532021
Estimating informativeness of samples with Smooth Unique Information
H Harutyunyan, A Achille, G Paolini, O Majumder, A Ravichandran, ...
International Conference on Learning Representations, 2021
282021
Supervision complexity and its role in knowledge distillation
H Harutyunyan, AS Rawat, AK Menon, S Kim, S Kumar
arXiv preprint arXiv:2301.12245, 2023
142023
Failure modes of domain generalization algorithms
T Galstyan, H Harutyunyan, H Khachatrian, GV Steeg, A Galstyan
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern …, 2022
142022
A meta-learning approach to predicting performance and data requirements
A Jain, G Swaminathan, P Favaro, H Yang, A Ravichandran, ...
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern …, 2023
92023
Fast structure learning with modular regularization
G Ver Steeg, H Harutyunyan, D Moyer, A Galstyan
Advances in Neural Information Processing Systems 32, 2019
82019
Disentangled representations via synergy minimization
G Ver Steeg, R Brekelmans, H Harutyunyan, A Galstyan
2017 55th Annual Allerton Conference on Communication, Control, and …, 2017
62017
Relaxed recursive transformers: Effective parameter sharing with layer-wise lora
S Bae, A Fisch, H Harutyunyan, Z Ji, S Kim, T Schuster
arXiv preprint arXiv:2410.20672, 2024
42024
Formal limitations of sample-wise information-theoretic generalization bounds
H Harutyunyan, G Ver Steeg, A Galstyan
2022 IEEE Information Theory Workshop (ITW), 440-445, 2022
42022
Identifying geopolitical event precursors using attention-based LSTMs
KSMT Hossain, H Harutyunyan, Y Ning, B Kennedy, N Ramakrishnan, ...
Frontiers in Artificial Intelligence 5, 893875, 2022
42022
Efficient covariance estimation from temporal data
H Harutyunyan, D Moyer, H Khachatrian, GV Steeg, A Galstyan
arXiv preprint arXiv:1905.13276, 2019
42019
Combining cnn and rnn for spoken language identification
H Harutyunyan, H Khachatrian
YerevaNN Blog on neural networks 11, 2016
42016
A little help goes a long way: Efficient llm training by leveraging small lms
AS Rawat, V Sadhanala, A Rostamizadeh, A Chakrabarti, W Jitkrittum, ...
arXiv preprint arXiv:2410.18779, 2024
22024
Mimetic Initialization Helps State Space Models Learn to Recall
A Trockman, H Harutyunyan, JZ Kolter, S Kumar, S Bhojanapalli
arXiv preprint arXiv:2410.11135, 2024
22024
Identifying and disentangling spurious features in pretrained image representations
R Darbinyan, H Harutyunyan, AH Markosyan, H Khachatrian
arXiv preprint arXiv:2306.12673, 2023
22023
In-context learning in presence of spurious correlations
H Harutyunyan, R Darbinyan, S Karapetyan, H Khachatrian
arXiv preprint arXiv:2410.03140, 2024
2024
On information captured by neural networks: connections with memorization and generalization
H Harutyunyan
arXiv preprint arXiv:2306.15918, 2023
2023
Systém momentálně nemůže danou operaci provést. Zkuste to znovu později.
Články 1–20