Yann LeCun

Research Statement

AI generated · 1d ago

Yann LeCun’s indexed work centers on the foundational architectures and training methods of modern deep learning. His most-cited contribution, the 2015 review “Deep learning,” synthesizes the field’s core principles and has become a canonical reference with over 85,000 citations. Earlier lead-author work established the technical backbone of neural networks, including backpropagation applied to handwritten zip code recognition and convolutional networks for images, speech, and time series, which together defined the CNN paradigm that dominates computer vision. His senior-authored research extends this framework to new domains, notably character-level convolutional networks for text classification and spectral networks on graphs, the latter seeding the field of geometric deep learning. These works collectively trace a consistent arc from supervised pattern recognition to non-Euclidean data, with citation counts in the thousands underscoring their sustained influence. Beyond his lead-author output, LeCun contributed to broader surveys and early architectures that shaped the field. He is a co-author of “Geometric Deep Learning: Going beyond Euclidean data,” a widely cited synthesis of graph- and manifold-based learning, and of “Signature verification using a Siamese time delay neural network,” an early demonstration of metric learning for verification tasks. His earlier work on “Optimal Brain Damage” and “Efficient BackProp” remains influential for network pruning and optimization practice. While no standalone software or datasets are indexed, the trajectory of these publications—from zip-code recognition to graph-structured data—reveals a sustained effort to generalize neural networks beyond their original Euclidean, supervised settings, establishing principles that continue to underpin contemporary AI research.

Most Cited Works

L’apprentissage profond, une révolution en intelligence artificielle

Yann LeCun

La lettre du Collège de France2016

18Citations

Informatique et sciences numériques

Yann LeCun

L’annuaire du Collège de France2018

Research Areas

Neural Networks and ApplicationsDomain Adaptation and Few-Shot LearningAdvanced Neural Network ApplicationsAdvanced Image and Video Retrieval TechnGenerative Adversarial Networks and Imag

Affiliations & Education

Education

1983–1987
Sorbonne Université(FR)
PhD · Computer Science
1978–1983
ESIEE(FR)
MS · Electrical Engineering

Employment

2003–present· now
New York University(US)
Professor · Courant Institute of Mathematical Sciences

Loading ORCID activity...

Sources

Research Statement

Most Cited Works

L’apprentissage profond, une révolution en intelligence artificielle

Informatique et sciences numériques

Research Areas

Affiliations & Education

Education

Employment