Yann LeCun

Developed convolutional networks and demonstrated them at scale on real recognition tasks.

Biography

A French-American computer scientist who developed convolutional neural networks and applied them to handwritten character recognition, producing systems deployed commercially for reading cheques well before the deep learning revival of the 2010s.

Key contributions

• Convolutional network architectures using weight sharing and local receptive fields.
• Early large-scale deployment of neural networks for handwriting recognition.
• The MNIST dataset, long the standard benchmark for image classification.

Impact

Weight sharing and locality, the two ideas at the core of convolutional networks, remain the reason vision models generalize from far less data than fully connected networks would require.