Number Sense Evolves in a Deep Neural Network (1 of 1) (IMAGE)

American Association for the Advancement of Science (AAAS)

Number Sense Evolves in a Deep Neural Network (1 of 1)

Caption

Artificial neurons that are tuned to preferred numbers of dots (such as one, two, four, and eight) spontaneously emerge in a neural deep-learning network that was merely trained to recognize objects in pictures. This material relates to a paper that appeared in the May 8th, 2019, issue of Science Advances, published by AAAS. The paper, by K. Nasr at University of Tübingen in Tübingen, Germany, and colleagues was titled, "Number detectors spontaneously emerge in a deep neural network designed for visual object recognition."

Credit

Andreas Nieder

Usage Restrictions

Please cite the owner of the material when publishing. This material may be freely used by reporters as part of news coverage, with proper attribution. Non-reporters must contact <i>Science</i> for permission.

License

Licensed content

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.