
Unsupervised Image Matching and Object Discovery as Optimization
Learning with complete or partial supervision is powerful but relies on ...
read it

A Spectral Regularizer for Unsupervised Disentanglement
Generative models that learn to associate variations in the output along...
read it

Learning about an exponential amount of conditional distributions
We introduce the Neural Conditioner (NC), a selfsupervised machine able...
read it

ModelPredictive Policy Learning with Uncertainty Regularization for Driving in Dense Traffic
Learning a policy using only observational data is challenging because t...
read it

Towards Understanding the Role of OverParametrization in Generalization of Neural Networks
Despite existing work on ensuring generalization of neural networks in t...
read it

Audio Source Separation with Discriminative Scattering Networks
In this report we describe an ongoing line of research for solving singl...
read it

Fast Incremental Learning for OffRoad Robot Navigation
A promising approach to autonomous driving is machine learning. In such ...
read it

Saturating AutoEncoders
We introduce a simple new regularizer for autoencoders whose hiddenuni...
read it

Fast Approximation of Rotations and Hessians matrices
A new method to represent and approximate rotation matrices is introduce...
read it

Universum Prescription: Regularization using Unlabeled Data
This paper shows that simply prescribing "none of the above" labels to u...
read it

Exploiting Linear Structure Within Convolutional Networks for Efficient Evaluation
We present techniques for speeding up the testtime evaluation of large ...
read it

Scene Parsing with Multiscale Feature Learning, Purity Trees, and Optimal Covers
Scene parsing, or semantic segmentation, consists in labeling each pixel...
read it

Fast approximations to structured sparse coding and applications to object classification
We describe a method for fast approximation of sparse coding. The input ...
read it

Eigenvalues of the Hessian in Deep Learning: Singularity and Beyond
We look at the eigenvalues of the Hessian of a loss function before and ...
read it

SuperResolution with Deep Convolutional Sufficient Statistics
Inverse problems in image and audio, and superresolution in particular,...
read it

Understanding Deep Architectures using a Recursive Convolutional Network
A key challenge in designing convolutional network models is sizing them...
read it

Differentially and nondifferentiallyprivate random decision trees
We consider supervised learning with random decision trees, where the tr...
read it

High Performance Computer Acoustic Data Accelerator: A New System for Exploring Marine Mammal Acoustics for Big Data Applications
This paper presents a new software model designed for distributed sonic ...
read it

The Loss Surfaces of Multilayer Networks
We study the connection between the highly nonconvex loss function of a...
read it

No More Pesky Learning Rates
The performance of stochastic gradient descent (SGD) depends critically ...
read it

Binary embeddings with structured hashed projections
We consider the hashing mechanism for constructing binary embeddings, th...
read it

Universal halting times in optimization and machine learning
The authors present empirical distributions for the halting time (measur...
read it

Prediction Under Uncertainty with ErrorEncoding Networks
In this work we introduce a new framework for performing temporal predic...
read it

A Closer Look at Spatiotemporal Convolutions for Action Recognition
In this paper we discuss several forms of spatiotemporal convolutions fo...
read it

Tunable Efficient Unitary Neural Networks (EUNN) and their application to RNNs
Using unitary (instead of general) matrices in artificial neural network...
read it

Hierarchical loss for classification
Failing to distinguish between a sheepdog and a skyscraper should be wor...
read it

ModelBased Planning in Discrete Action Spaces
Planning actions using learned and differentiable forward models of the ...
read it

Fast Convolutional Nets With fbfft: A GPU Performance Evaluation
We examine the performance profile of Convolutional Neural Network train...
read it

Emergence of ComplexLike Cells in a Temporal Product Network with Local Receptive Fields
We introduce a new neural architecture and an unsupervised algorithm for...
read it

Disentangling factors of variation in deep representations using adversarial training
We introduce a conditional generative model for learning to disentangle ...
read it

EntropySGD: Biasing Gradient Descent Into Wide Valleys
This paper proposes a new optimization algorithm called EntropySGD for ...
read it

Predicting Deeper into the Future of Semantic Segmentation
The ability to predict and therefore to anticipate the future is an impo...
read it

Energybased Generative Adversarial Network
We introduce the "Energybased Generative Adversarial Network" model (EB...
read it

Geometric deep learning: going beyond Euclidean data
Many scientific fields study data with an underlying structure that is a...
read it

Recurrent Orthogonal Networks and LongMemory Tasks
Although RNNs have been shown to be powerful tools for processing sequen...
read it

Deep multiscale video prediction beyond mean square error
Learning to predict future images from a video sequence involves the con...
read it

Stacked WhatWhere Autoencoders
We present a novel architecture, the "stacked whatwhere autoencoders" ...
read it

What is the Best Feature Learning Procedure in Hierarchical Recognition Architectures?
(This paper was written in November 2011 and never published. It is post...
read it

A mathematical motivation for complexvalued convolutional networks
A complexvalued convolutional network (convnet) implements the repeated...
read it

Learning Stable Group Invariant Representations with Convolutional Networks
Transformation groups, such as translations or rotations, effectively ex...
read it

Deep learning with Elastic Averaging SGD
We study the problem of stochastic optimization for deep learning in the...
read it

Explorations on high dimensional landscapes
Finding minima of a real valued nonconvex function over a high dimensio...
read it

Stereo Matching by Training a Convolutional Neural Network to Compare Image Patches
We present a method for extracting depth information from a rectified im...
read it

Signal Recovery from Pooling Representations
In this work we compute lower Lipschitz bounds of ℓ_p pooling operators ...
read it

Deep Convolutional Networks on GraphStructured Data
Deep Learning's recent successes have mostly relied on Convolutional Net...
read it

Learning to Linearize Under Uncertainty
Training deep feature hierarchies to solve supervised learning tasks has...
read it

Adaptive learning rates and parallelization for stochastic, sparse, nonsmooth gradients
Recent work has established an empirically successful framework for adap...
read it

Unsupervised Feature Learning from Temporal Data
Current stateoftheart classification and detection algorithms rely on...
read it

Unsupervised Learning of Spatiotemporally Coherent Metrics
Current stateoftheart classification and detection algorithms rely on...
read it

Efficient Object Localization Using Convolutional Networks
Recent stateoftheart performance on humanbody pose estimation has be...
read it
Yann LeCun
is this you? claim profile
Yann LeCun is a French computer scientist who works primarily in the machine learning, computer vision, mobile robotics and computer neuroscience fields. He is a Silver Professor at the Mathematical Sciences Courant Institute, New York University, Vice President, Chief AI Scientist, Facebook. He is well known for working on the recognition of optical character and computer vision through convolutionary neural networks and is the founder of convolutionary networks. He is also one of the main developers of DjVu compression technology. With Léon Bottou he codeveloped the Lush programming language.
He is cowinner of the ACM A.M. Turing Award 2018 for his deep education work.
Yann LeCun was born in 1960 in the Paris suburbs of SoisysousMontmorency. He graduated with a Diplôme d’Ingénieur from the ESIEE Paris in 1983 and a PhD in computer science from the Pierre and Marie Curie University in 1987 during which he proposed an early form of the backpropagation learning algoritude for neural networks.
In 1988, Lawrence D. Jackel joined the Adaptive Systems Research Department at AI&T Bell Laboratories, Holmdel, New Jersey, USA, where he developed several new methods of machine learning like the bioinspired picture recognition model called Convolutional Neural Networks, the methods of “optimal brain damage” and m Graph Transformer Networks. The system of bank check recognition that NCR and other companies were used to help with was more than 10% of the checking in the USA in the late 1990’s and early 2000’s. In 1996, he joined AT&T LabsResearch as Head of the Department for Image Processing Research, part of the Lawrence Rabiner Speech and Image Processing Research Laboratory. Léon Bottou and Vladimir Vapnik are his collaborators at AT&T.
He joined New York University in 2003 after having worked as a fellow of the NEC Research Institute in Princeton, NJ, where he is Silver Professor of Computer Science of Neural Science at the Courant Institute of Mathematical Science and the Center for NearKnowledge. He is also a professor at the Engineering School of Tandon. At NYU, he primarily worked on energybased models in supervised and unchecked education, computer vision feature learning and mobile robotics.In 2012, he was appointed as the founding director of the NYU Center for Data Science. LeCun became the first Facebook AI Research Director in New York City on December 9, 2013, and retired from the NYUCDS Directorship in early 2014.
In 2013, he and Yoshua Bengio cofounded the International Learning Conference that adopted an Open Review process that he previously recommended on his website. He was chairman and organizer of the “Learning Workshop,” organized each year in Snowbird, Utah, from 1986 to 2012. He is a member of the Institute of Pure and Applied Mathematics’ Science Advisory Board at UCLA. He is CoDirector, CIFAR’s Research Program for Learning in Machinery and Brain. In 2016, he was the visiting IT professor in the “Chaire Annuelle Informatique et Sciences Numériques” at the Collège de France in Paris. His “inaugural lesson” was an important event in the intellectual life of Paris in 2016.
LeCun is a member of the National Academy of Engineering of the United States, the recipient of the 2014 Pioneer Award for IEEE Neural Network and the 2015 PAMI Distinguished Researcher Award. In 2017, LeCun rejected an invitation to lecture at Saudi Arabia’s King Abdullah University for his belief that he was a terrorist in that country because of his atheism.
He received the Harold Pender Award from the University of Pennsylvania in September 2018. He was awarded the doctorate Honoris Causa from EPFL in October 2018.
LeCun won the Turing award in March 2019, sharing it with Yoshua Bengio and Geoffrey Hinton.