Home About the PI Publications Preprints Activities Teaching Group Apply

SGD and Weight Decay Secretly Minimize the Rank of Your Neural Network
T. Galanti, Z. Siegel, A. Gupte, T. Poggio
Conference on Parsimony and Learning, CPAL, 2025
NeurIPS Workshop on Mathematics of Modern Machine Learning, M3L, 2024

Distributed Speculative Inference (DSI): Speculation Parallelism for Provably Faster Lossless Language Model Inference
N. Timor, J. Mamou, D. Korat, M. Berchansky, O. Pereg, M. Wasserblat, T. Galanti, M. Gordon, D. Harel
International Conference on Learning Representations, ICLR, 2025

Formation of Representations in Neural Networks
L. Ziyin, I. L. Chuang, T. Galanti, T. Poggio
International Conference on Learning Representations, ICLR, 2025
Spotlight presentation (5% acceptance)

On the Power of Decision Trees in Auto-Regressive Language Modeling
Y. Gan, T. Galanti, T. Poggio, E. Malach
Neural Information Processing Systems, NeurIPS, 2024

Distributed Speculative Inference of Large Language Models is Provably Faster
N. Timor, J. Mamou, D. Korat, M. Berchansky, O. Pereg, M. Wasserblat, T. Galanti, M. Gordon, D. Harel
NeurIPS Workshop on Efficient Natural Language and Speech Processing (ENLSP), 2024 (published at PMLR).

Norm-Based Generalization Bounds for Sparse Neural Networks
T. Galanti, M. Xu, L. Galanti, T. Poggio
Neural Information Processing Systems, NeurIPS, 2023

Reverse Engineering Self-Supervised Learning
I. Ben-Shaul, R. Shwartz-Ziv*, T. Galanti*, S. Dekel, Y. LeCun
Neural Information Processing Systems, NeurIPS, 2023

Comparative Generalization Bounds for Deep Neural Networks
T. Galanti, L. Galanti, I. Ben-Shaul
Transactions in Machine Learning Research, TMLR, 2023

Exploring the Approximation Capabilities of Multiplicative Neural Networks for Smooth Functions
I. Ben-Shaul, T. Galanti, S. Dekel
Transactions in Machine Learning Research, TMLR, 2023

Feature Learning in Deep Classifiers Through Intermediate Neural Collapse
A. Rangamani, M. Lindegaard, T. Galanti, T. Poggio
International Conference on Machine Learning, ICML, 2023

Dynamics of Deep Classifiers Trained with the Square Loss: Normalization, Low Rank, Neural Collapse and Generalization Bounds
M. Xu, A. Rangamani, Q. Liao, T. Galanti, T. Poggio
Research (a Science partner journal), 2023

Image2Point: 3D Point-Cloud Understanding with 2D Image Pretrained Models
C. Xu, S. Yang, T. Galanti, B. Wu, X. Yue, B. Zhai, W. Zhan, P. Vajda, K. Keutzer, M. Tomizuka
IEEE European Conference on Computer Vision, ECCV, 2022

Improved Generalization Bounds for Transfer Learning via Neural Collapse
T. Galanti, A. Gyorgy, M. Hutter
ICML Workshop on Pretraining: Perspectives, Pitfalls and Paths Forward, 2022

On the Implicit Bias Towards Depth Minimization in Deep Neural Networks
T. Galanti, L. Galanti, I. Ben-Shaul
Conference on the Mathematical Theory of Deep Neural Networks, DEEPMATH, 2022
Workshop on the Theory of Overparameterized Machine Learning, TOPML, 2022

On the Role of Neural Collapse in Transfer Learning
T. Galanti, A. Gyorgy, M. Hutter
International Conference on Learning Representations, ICLR, 2022

Weakly Supervised Discovery of Semantic Attributes
A. A. Ali, T. Galanti, E. Zheltonozhskii, C. Baskin, L. Wolf
Causal Learning and Reasoning, CLeaR, 2022

Meta Internal Learning
R. Ben Sadoun, S. Gur, T. Galanti, L. Wolf
Neural Information Processing Systems, NeurIPS, 2021

On Random Kernels of Residual Architectures
E. Littwin*, T. Galanti*, L. Wolf
Uncertainty in Artificial Intelligence, UAI, 2021

Risk Bounds for Unsupervised Cross-Domain Mapping with IPMs
T. Galanti, S. Benaim, L. Wolf
Journal of Machine Learning Research, JMLR, 2021

Evaluation Metrics for Conditional Image Generation
Y. Benny, T. Galanti, S. Benaim, L. Wolf
International Journal of Computer Vision, IJCV, 2021

On the Modularity of Hypernetworks 
T. Galanti, L. Wolf
Neural Information Processing Systems, NeurIPS, 2020
Oral presentation (1% acceptance)

On Infinite-Width Hypernetworks
E. Littwin*, T. Galanti*, L. Wolf
Neural Information Processing Systems, NeurIPS, 2020

Domain Intersection and Domain Difference
S. Benaim, M. Khaitov, T. Galanti, L. Wolf
IEEE International Conference on Computer Vision, ICCV, 2019

Unsuperivsed Learning of the Set of Local Maxima
L. Wolf, S. Benaim, T. Galanti
International Conference on Learning Representations, ICLR, 2019

Emerging Disentanglement in Auto-Encoder Based Unsupervised Image Content Transfer
O. Press, T. Galanti, S. Benaim, L. Wolf
International Conference on Learning Representations, ICLR, 2019

A Formal Approach to Explainability
L. Wolf, T. Galanti, T. Hazan
Artificial Intelligence, Ethics and Society, AIES, 2019

Generalization Bounds for Cross-Domain Mapping with WGANs
T. Galanti, S. Benaim, L. Wolf
NeurIPS Workshop on Integration of Deep Learning Theories, 2018

Estimating the Success of Unsupervised Image to Image Translation
S. Benaim*, T. Galanti*, L. Wolf
IEEE European Conference on Computer Vision, ECCV, 2018

The Role of Minimal Complexity in Unsupervised Learning of Semantic Mappings
T. Galanti, L. Wolf, S. Benaim
International Conference on Learning Representations, ICLR, 2018

A Theoretical Framework for Deep Transfer Learning
T. Galanti, T. Hazan, L. Wolf
Information and Inference: A Journal of the IMA, IMAIAI, 2016