01 · A life in computing
The person behind the work
Yann LeCun developed convolutional neural networks into effective systems for recognizing visual patterns. His work at Bell Labs combined trainable feature extraction with end-to-end learning, including applications to handwritten character recognition. At New York University and Facebook he continued research on representation learning and machine intelligence. His contribution to the joint deep-learning award particularly connects neural-network architecture with the engineering needed for useful perception systems.
The contribution, explained
Yoshua Bengio, Geoffrey Hinton and Yann LeCun made conceptual and engineering advances that turned deep neural networks into a central tool of computing. Rather than hand-design every feature, a network can learn intermediate representations from data. Training adjusts its parameters so that many layers work together towards a task.
Why it matters
Learned representations helped transform tasks including image recognition, speech processing and other forms of pattern learning.