Publications
Google Scholar Profile
2024
Provable Posterior Sampling with Denoising Oracles through Tilted Transport, Han J, Bruna J, submitted
Benefits of Rank in Attention Layers, Amsel N, Yehudai G, Bruna J, submitted
How Truncating Weights Improves Reasoning in Language Models, Chen L, Bruna J, Bietti A, submitted.
Stochastic Optimal Control Matching Domingo-Enrich C, Han J, Amos B, Bruna J, Chen R, submitted.
Computational-Statistical Gaps in Gaussian Single-Index Models Damian A, Pillaud-Vivien L, Lee J, Bruna J, Colt 24
Symmetric Single-Index Learning Zweig A, Bruna J ICLR 24
2023
On Learning Gaussian Multi-Index Models with Gradient Flow Bietti A, Bruna J, Pillaud-Vivien L, submitted.
A neural Collapse perspective on feature evolution in graph neural networks Kothapalli V, Tirer T, Bruna J NeurIPS 23
Inverse Dynamics Pretraining Learns Good Representations for Multitask Imitation, Brandfonbrener D, Nachum O, Bruna J NeurIPS 23
On Single-Index Models beyond Gaussian Data Bruna J, Pillaud-Vivien L, Zweig A, NeurIPS 23
Conditionally strongly log-concave generative models Guth F, Lempereur E, Bruna J, Mallat S, ICML 23
Reliable coarse-grained turbulent simulations through combined offline learning and neural emulation Pedersen C, Perezhogin P, Zanna L, Bruna J, ICML 23 SynS ML Workshop
Data-driven multiscale modeling of subgrid parametrizations in climate models Otness K, Zanna L and Bruna J, ICLR 23 Climate and AI workshop (Best ML paper award)
On Gradient Descent Convergence beyond the Edge of Stability, Chen Lei and Bruna J, ICML 23
2022
A functional-space Mean-Field theory of partially-trained three-layer neural networks, Chen Z, Vanden-Eijnden E, Bruna, J, submitted.
Towards Antisymmetric Neural Ansatz Separation, Zweig, A, Bruna J, preprint.
Learning Single-Index Models with Shallow Neural Networks, Bietti A, Song MJ, Sanford C, Bruna J, NeurIPS 22.
On Non-Linear Operators for Geometric Deep Learning, Sergeant-Perthuis G, Maier J, Bruna J, Oyallon E, NeurIPS 22.
Exponential Separations in Symmetric Neural Networks, Zweig A and Bruna J, NeurIPS 22
When does return-conditioned supervised learning work for offline reinforcement learning?, Branfonbrener D, Bietti A, Buckman J, Laroche R, Bruna, J, NeurIPS 22.
Lattice-based methods surpass sum-of-squares in clustering Zadik I, Song MJ, Wein A and Bruna J, COLT
On Feature Learning in Neural Networks with Global Convergence Guarantees, Chen, Zhengdao, Vanden-Eijnden E, Bruna, J, ICLR
Neural Galerkin Scheme with Active Learning for High-dimensional Evolution Equations, Bruna, J, Peherstofer, B, Vanden-Eijnden, E, preprint.
Extended Unconstrained Features Models for exploring deep Neural Collapse, Tirer Tom and Bruna, Joan, ICML
Neural Fields as Learnable Kernels for 3d reconstruction, Williams F, Gojcic Z, Khamis, S, Zorin D, Bruna , Fidler S, Litany, O, CVPR
Cartoon Explanations of Image Classifiers, Kolek, S, Nguyen, D, Levie, R, Bruna, J, Kutyniok, G, ECCV (Oral)
2021
Dual Training of Energy-Based Models with Overparametrised Shallow Neural Networks, Domingo-Enrich, C, Bietti, A, Gabrie, M, Bruna, J, Vanden-Eijnden, E, submitted.
Offline Contextual Bandits with Overparametrised Models, Brandfonbrener, D, Whitney, W, Ranganath, R, Bruna, J, NeurIPS21 (Spotlight)
On the Sample Complexity of Learning under Invariance and Geometric Stability, Bietti, A, Venturi, L, Bruna, J, NeurIPS21
On the Cryptographic Hardness of Learning Single Periodic Neurons, Song, M-J, Zadik, I, Bruna, J, NeurIPS21
An Extensible Benchmark Suite for Learning to Simulate Physical Systems, Otness, K, Gjoka, A, Bruna, J, Panozzo, D, Peherstorfer, B, Schneider, T, Zorin, D, NeurIPS21 (Benchmark Datasets Track)
[Learning the Relevant Substructures for Tasks on Graph Data], Chen, L, Chen, Z, Bruna, J, ICASSP21.
Kernel-Based Smoothness Analysis of Residual Networks, Tirer T, Bruna J, Giryes, R., MSML21.
Neural Splines: Fitting 3d surfaces with Ininitely-Wide Neural Networks, Williams, F, Trager, M, Bruna, J and Zorin, D., CVPR21 (Oral).
On Energy-Based Models with Overparametrised Shallow Neural Networks, Domingo-Enrich, C, Bietti, A, Vanden-Eijnden, E, Bruna, J, ICML21.
A Functional Perspective on Learning Symmetric Functions with Neural Networks, Zweig, A, Bruna, J. ICML21.
Overfitting and Optimization in Offline Policy Learning, Brandfonbrener, D, Whitney, W, Ranganath R, and Bruna, J, ICML21.
On Graph Neural Networks versus Graph-Augmented MLPs, Chen, L, Chen, Z, Bruna, J, ICLR21
Depth Separation Beyond Radial Functions, Bruna, Joan, Jelassi, Samy, Ozuch, T and Venturi, Luca, submitted
2020
Adaptive Test Allocation for Outbreak Detection and Tracking in Social Contact Networks, Batlle, P, Bruna, J, Fernandez-Granda, C., Preciado, V, SIAM Journal on Control and Optimization.
Depth Separation for reduced Deep Networks in nonlinear model reduction: Distilling Shock Waves in nonlinear hyperbolic problems, Rim D, Venturi, L, Bruna J, Peherstofer, B., submitted.
On Sparsity in Overparametrised Shallow ReLU Networks, De Dios, J and Bruna, J., submitted.
A New Approach to Observational Cosmology using the Scattering Transform, Cheng, S, Ting, Y-S, Menard, B and Bruna Joan, Journal of the Royal Astronomical Society International Astrostatistics Association Outstanding Publication Award.
Continuous LWE, Bruna, J, Regev, O, Song, MJ, Tang, Y, STOC 2021.
A permutation-equivariant Neural Network architecture for auction design, Rahme, J, Jelassi, S., Bruna, J and Weinberg, M, AAAI21.
Stability Properties of Graph Neural Networks, Gama F, Bruna, J, Ribeiro, A, IEEE Transactions on Signal Processing.
A Dynamical Central Limit Theorem for Shallow Neural Networks, Chen, Z, Rotskoff, G, Bruna, J, Vanden-Eijnden, E. NeurIPS 20.
Can Graph Neural Networks Count Substructures?, Chen, Z., Chen, L, Villar, S, Bruna, J, NeurIPS.
A mean-field analysis of two-player zero-sum games, Domingo-Enrich, C., Jelassi, S., Mensch, A, Rotskoff, G, Bruna, J, NeurIPS.
IDEAL: Inexact Decentralized Accelerated Augmented Lagrangian Method, Arjevani, Y. Bruna, J., Can B, Gurbuzbalaban M, Jegelka S, Lin H., NeurIPS 20 (spotlight).
In-Distribution Interpretability for Challenging Modalities, Heiss, C, Levie, R, Resnick, C, Kutyniok G, and Bruna, J, ICML Interpretability Workshop.
[Stability of Graph Neural Networks to Relative Perturbations], Gama, F, Bruna, J, Ribeiro, A, ICASSP 2020.
Provably Efficient Third-Person Imitation from Offline Obsevation, Zweig, A, Bruna J, UAI 2020.
Extra-Gradient with Player Sampling for Provable Fast Convergence in n-player Games, Domingo C, Jelassi, S, Scieur, D, Mensch, A, Bruna J, ICML 2020 (previous version appeared in Bridging Game Theory and Deep Learning Workshop NeurIPS 2019).
2019
Probing the State of the Art: A Critical Look at Visual Representation Evaluation, Resnick, C., Zhan, Z., Bruna, J., preprint.
The Scattering Representation, Bruna J, book chapter, part of forthcoming Mathematics of Deep Learning, Cambridge Univ. Press
Pure and Spurious Critical Points: a Geometric Study of Linear Networks, Trager, M, Kohn, K, Bruna, J, ICLR 2020
Geometric Insights into the Convergence of Nonlinear TD Learning, Brandfonbrener D, Bruna J, ICLR 2020, NeurIPS 2019 Workshop on Optimization Foundation for Reinforcement Learning
Stability Properties of Graph Neural Networks, Gama, F, Bruna J, Ribeiro A , submitted
Smaller Embeddings for Large Scale Knowledge Base Completion, Lacroix, T. Obozinski, G., Bruna J, Usunier, N., submitted.
Attributed Random Walk as Matrix Factorization, Chen L, Gong, S, Bronstein, M, Bruna, J, NeurIPS 2019 Graph Representation Learning Workshop
Spurious Valleys in One-hidden-layer Neural Network Optimization Landscapes, Venturi, L. Bandeira, A. Bruna, J., JMLR’19.
Gradient Dynamics of Univariate ReLU Networks, Williams, F. Trager, M. Panozzo, D. Zorin, D, Bruna J, NeurIPS’19.
Finding the Needle in the Haystack with Convolutions: on the benefits of architectural Bias, d’Ascoli, S, Sagun, L, Bruna J, Biroli, G., NeurIPS’19
Stability of Graph Scattering Transforms, Gama F, Bruna J, Ribeiro, A, NeurIPS’19
On the Equivalence between Graph Isomorpmishm Testing and Function Approximation with GNNs, Chen, Z., Villar, S, Chen, L, Bruna J, NeurIPS’19
On the Expressive Power of Deep Polynomial Neural Networks, Kileel J, Trager M, Bruna J, NeurIPS’19
Approximating Unitary Matrices with Effective Givens Factorization, Frerix, Th., Bruna J., ICML 2019
Global Convergence of Neuron Birth-Death Dynamics, Rotskoff, G., Jelassi, S., Bruna J. Vanden-Eijnden, E., ICML 2019
Geometric Deep Prior for Surface Reconstruction, Williams, F. Schneider, T. Silva C, Zorin, D. Bruna, J and Panozzo, D., CVPR 2019
Diffusion Scattering Transforms on Graphs, Gama, F. Ribeiro, A. Bruna, J., ICLR 19
Supervised Community Detection with Non-backtracking Graph Neural Networks, Chen, Z. Li, Xiang, Bruna, J., ICLR 19
Multiscale Sparse Microcanonical Models, Bruna, J., Mallat, S, Mathematical Statistics and Learning, 2019
2018
Graph Neural Networks for IceCube Signal Classification, Choma, N. Monti, F. Gerhardt, L., Palczewski, T, Ronaghi, Z, Prabhat, Bhimji, W., Bronstein, M., Klein, S., Bruna, J., ICMLA (Oral) 2018, Best Paper Award
Pommerman: A multi-agent playground, Resnick, C., Eldridge, W. Ha, D, Britz, D., Foerster, J., Togelius, J. Cho, K. and Bruna, J., AIIDE Workshop (Oral) 2018
Backplay: “Man muss immer umkehren”, Resnick, C. Raileanu, R. , Kapoor, S. Peysakhovich, A. Cho, K, Bruna J., arxiv,18
Planning with Arithmetic and Geometric Attributes, Folque, D. Sukhbaatar, S. , Szlam, A, Bruna, J., ICML Workshop 18
Surface Networks, Kostrikov,I. Panozzo,D., Zorin, D and Bruna,J, CVPR’18 (oral)
Divide and Conquer Networks, Nowak, A., Folque, D Bruna, J., ICLR 2018
Few-Shot Learning with Graph Neural Networks, Garcia, V. and Bruna, J., ICLR 2018
2017
Neural Message Passing for Jet Physics, Henrion, I. Brehmer, J. Bruna, J., Cho, K., Cranmer, K., Louppe, G, Rochette, G. NeurIPS’17 Deep Learning for Physical Sciences Workshop
Mathematics of Deep Learning, R. Vidal, Bruna, J, Giryes, R. and Bruna, J. CDC 2017
A Note on Learning Algorithms for Quadratic Assignment with Graph Neural Networks, Nowak, A., Villar, S., Bandeira, A. and Bruna, J. ICML’17 PADL Workshop
Topology and Geometry of Half-Rectified Networks Optimization, Freeman, D. , Bruna, J. ICLR 2017
Geometric Deep Learning: going beyond Euclidean Data, Bronstein, M. Bruna, J., Szlam, A., LeCun, Y. and Vandergyst, P. IEEE Sig. Proc. Magazine, 2017
Inverse Problems with Invariant Multiscale Statistics, Dokmanic, I. Bruna, J., Mallat, S. and De Hoop, M. SPARS 2017 (oral)
Understanding Trainable Sparse Coding via Matrix Factorization, Moreau, Th., Bruna, J. ICLR 2017
2016
Super-Resolution with Deep Convolutional Sufficient Statistics, Bruna, J., Sprechmann, P. and LeCun, Y. ICLR 2016
Deep Convolutional Networks on Graph-Structured Data, Henaff, M, Bruna, J. and LeCun, Y. arxiv
2015
Unsupervised Learning of Spatiotemporally Coherent Metrics, Goroshin, R, Bruna, J., Thomson, J. Eigen, D and Lecun, Y, ICCV, 2015
A theoretical argument for complex-valued convolutional networks, Bruna, J, Chintala, S., Piantino, S., Szlam, A. and Tygert, M, Neural Computation, 2015
Video (language) modeling: a baseline for generative models of natural videos, Ranzato, M.A. , Szlam, A., Bruna, J., Mathieu, M., Collobert, R. and Chopra, S. Preprint, 2015
Training Convolutional Networks with Noisy Labels, Sukhbaatar, S., Bruna, J., Paluri, M. Bourdev, L. Fergus, R, ICLR Workshop, 2015
Source Separation with Scattering Non-negative Matrix Factorization, Bruna, J., Sprechmann, P. and LeCun, Y. ICASSP 2015
2014
Exploiting Linear Structure within Convolutional Networks for Efficient Evaluation, Denton, E., Zaremba, W., Bruna, J., LeCun, Y. and Fergus, R. NeurIPS 2014
Intriguing Properties of Neural Networks, Szegedy, C., Zaremba, W., Sutskever, I., Bruna, J., Erhan, D., Goodfellow, I. and Fergus, R., ICLR 2014
Spectral Networks and Locally Connected Networks on Graphs, Bruna, J., Zaremba, W., Szlam, A. and LeCun, Y., ICLR 2014
Signal Recovery from Pooling Representations, Bruna, J., Szlam, A. and LeCun, Y. ICML 2014
Intermittent Process Analysis with Scattering Moments, Bruna, J., Mallat, S. Bacry, E. and Muzy, J-F. Annals of Statistics, 2014
2013
Blind Deconvolution with Re-weighted Sparsity Promotion, Krishnan, D., Bruna, J. and Fergus, R. arxiv 2013
Audio Texture Synthesis with Scattering Moments, Bruna, J., Mallat, S. Preprint, 2013
Learning Stable Group Invariant Representations with Convolutional Networks, Bruna, J., LeCun, Y., Szlam, A. ICLR Workshop, 2013
2012
Scattering Representations for Recognition, Bruna, J. PhD Thesis, 2012
Invariant Scattering Convolutional Networks, Bruna, J., Mallat, S., IEEE trans of PAMI, 2012
2011
Classification with Scattering Operators, Bruna, J., Mallat, S., CVPR, 2011
Classification with Invariant Scattering Representations, Bruna, J, Mallat, S. IEEE IVMSP 2011
2010
- Geometric Models with Co-occurrence Groups, Bruna, J., Mallat, S. ESANN, 2010