Get our free extension to see links to code for papers anywhere online!

Chrome logo Add to Chrome

Firefox logo Add to Firefox

Picture for Tomaso Poggio

Distribution of Classification Margins: Are All Data Equal?


Jul 21, 2021
Andrzej Banburski, Fernanda De La Torre, Nishka Pant, Ishana Shastri, Tomaso Poggio

* Previously online as CBMM Memo 115 on the CBMM MIT site 

  Access Paper or Ask Questions

The Effects of Image Distribution and Task on Adversarial Robustness


Feb 21, 2021
Owen Kunhardt, Arturo Deza, Tomaso Poggio

* Under review at ICML 2021 

  Access Paper or Ask Questions

Explicit regularization and implicit bias in deep network classifiers trained with the square loss


Dec 31, 2020
Tomaso Poggio, Qianli Liao


  Access Paper or Ask Questions

CUDA-Optimized real-time rendering of a Foveated Visual System


Dec 15, 2020
Elian Malkin, Arturo Deza, Tomaso Poggio

* 16 pages, 13 figures, presented at the Shared Visual Representations in Human and Machine Intelligence Workshop (SVRHM NeurIPS 2020) 

  Access Paper or Ask Questions

Biologically Inspired Mechanisms for Adversarial Robustness


Jun 29, 2020
Manish V. Reddy, Andrzej Banburski, Nishka Pant, Tomaso Poggio

* 25 pages, 15 figures 

  Access Paper or Ask Questions

Hierarchically Local Tasks and Deep Convolutional Networks


Jun 29, 2020
Arturo Deza, Qianli Liao, Andrzej Banburski, Tomaso Poggio

* A pre-print. Submitted to the Conference of Neural Information Processing Systems (NeurIPS) 2020 

  Access Paper or Ask Questions

For interpolating kernel machines, the minimum norm ERM solution is the most stable


Jun 28, 2020
Akshay Rangamani, Lorenzo Rosasco, Tomaso Poggio


  Access Paper or Ask Questions

Double descent in the condition number


Dec 12, 2019
Tomaso Poggio, Gil Kur, Andrzej Banburski


  Access Paper or Ask Questions

Theoretical Issues in Deep Networks: Approximation, Optimization and Generalization


Aug 25, 2019
Tomaso Poggio, Andrzej Banburski, Qianli Liao

* arXiv admin note: text overlap with arXiv:1611.00740 

  Access Paper or Ask Questions

Theory III: Dynamics and Generalization in Deep Networks - a simple solution


Apr 11, 2019
Andrzej Banburski, Qianli Liao, Brando Miranda, Lorenzo Rosasco, Bob Liang, Jack Hidary, Tomaso Poggio

* 50 pages, 11 figures. This replaces previous versions of Theory III, that appeared on Arxiv [arXiv:1806.11379, arXiv:1801.00173] or on the CBMM site. v2: Some arguments in sections 3, 7 and the appendix have been strengthened, in particular an observation on intrinsic normalization of standard gradient descent 

  Access Paper or Ask Questions

Biologically-plausible learning algorithms can scale to large datasets


Nov 25, 2018
Will Xiao, Honglin Chen, Qianli Liao, Tomaso Poggio


  Access Paper or Ask Questions

An analysis of training and generalization errors in shallow and deep networks


Aug 21, 2018
Hrushikesh Mhaskar, Tomaso Poggio


  Access Paper or Ask Questions

A Surprising Linear Relationship Predicts Test Performance in Deep Networks


Jul 25, 2018
Qianli Liao, Brando Miranda, Andrzej Banburski, Jack Hidary, Tomaso Poggio


  Access Paper or Ask Questions

Theory IIIb: Generalization in Deep Networks


Jun 29, 2018
Tomaso Poggio, Qianli Liao, Brando Miranda, Andrzej Banburski, Xavier Boix, Jack Hidary

* 38 pages, 7 figures 

  Access Paper or Ask Questions

Approximate inference with Wasserstein gradient flows


Jun 12, 2018
Charlie Frogner, Tomaso Poggio


  Access Paper or Ask Questions

Theory of Deep Learning III: explaining the non-overfitting puzzle


Jan 16, 2018
Tomaso Poggio, Kenji Kawaguchi, Qianli Liao, Brando Miranda, Lorenzo Rosasco, Xavier Boix, Jack Hidary, Hrushikesh Mhaskar


  Access Paper or Ask Questions

Theory of Deep Learning IIb: Optimization Properties of SGD


Jan 07, 2018
Chiyuan Zhang, Qianli Liao, Alexander Rakhlin, Brando Miranda, Noah Golowich, Tomaso Poggio


  Access Paper or Ask Questions

Fisher-Rao Metric, Geometry, and Complexity of Neural Networks


Nov 05, 2017
Tengyuan Liang, Tomaso Poggio, Alexander Rakhlin, James Stokes

* 31 pages, 7 figures 

  Access Paper or Ask Questions

Pruning Convolutional Neural Networks for Image Instance Retrieval


Jul 18, 2017
Gaurav Manek, Jie Lin, Vijay Chandrasekhar, Lingyu Duan, Sateesh Giduthuri, Xiaoli Li, Tomaso Poggio

* 5 pages 

  Access Paper or Ask Questions

Do Deep Neural Networks Suffer from Crowding?


Jun 26, 2017
Anna Volokitin, Gemma Roig, Tomaso Poggio

* CBMM memo 

  Access Paper or Ask Questions

Theory II: Landscape of the Empirical Risk in Deep Learning


Jun 22, 2017
Qianli Liao, Tomaso Poggio

* Merged figures to make the main text more compact. Moved some similar figures to the appendix 

  Access Paper or Ask Questions

Why and When Can Deep -- but Not Shallow -- Networks Avoid the Curse of Dimensionality: a Review


Feb 04, 2017
Tomaso Poggio, Hrushikesh Mhaskar, Lorenzo Rosasco, Brando Miranda, Qianli Liao


  Access Paper or Ask Questions

Compression of Deep Neural Networks for Image Instance Retrieval


Jan 18, 2017
Vijay Chandrasekhar, Jie Lin, Qianli Liao, Olivier Morère, Antoine Veillard, Lingyu Duan, Tomaso Poggio

* 10 pages, accepted by DCC 2017 

  Access Paper or Ask Questions

Streaming Normalization: Towards Simpler and More Biologically-plausible Normalizations for Online and Recurrent Learning


Oct 19, 2016
Qianli Liao, Kenji Kawaguchi, Tomaso Poggio


  Access Paper or Ask Questions

Deep vs. shallow networks : An approximation theory perspective


Aug 10, 2016
Hrushikesh Mhaskar, Tomaso Poggio

* 14 pages, 4 figures, to be published in a Journal 

  Access Paper or Ask Questions

View-tolerant face recognition and Hebbian learning imply mirror-symmetric neural tuning to head orientation


Jun 05, 2016
Joel Z. Leibo, Qianli Liao, Winrich Freiwald, Fabio Anselmi, Tomaso Poggio


  Access Paper or Ask Questions

Learning Functions: When Is Deep Better Than Shallow


May 29, 2016
Hrushikesh Mhaskar, Qianli Liao, Tomaso Poggio


  Access Paper or Ask Questions

Nested Invariance Pooling and RBM Hashing for Image Instance Retrieval


Apr 14, 2016
Olivier Morère, Jie Lin, Antoine Veillard, Vijay Chandrasekhar, Tomaso Poggio

* Image Instance Retrieval, CNN, Invariant Representation, Hashing, Unsupervised Learning, Regularization. arXiv admin note: text overlap with arXiv:1601.02093 

  Access Paper or Ask Questions

Bridging the Gaps Between Residual Learning, Recurrent Neural Networks and Visual Cortex


Apr 13, 2016
Qianli Liao, Tomaso Poggio


  Access Paper or Ask Questions