Human-level concept learning through probabilistic program induction.

Journal: Science (New York, N.Y.)
Published Date:

Abstract

People learning new concepts can often generalize successfully from just a single example, yet machine learning algorithms typically require tens or hundreds of examples to perform with similar accuracy. People can also use learned concepts in richer ways than conventional algorithms-for action, imagination, and explanation. We present a computational model that captures these human learning abilities for a large class of simple visual concepts: handwritten characters from the world's alphabets. The model represents concepts as simple programs that best explain observed examples under a Bayesian criterion. On a challenging one-shot classification task, the model achieves human-level performance while outperforming recent deep learning approaches. We also present several "visual Turing tests" probing the model's creative generalization abilities, which in many cases are indistinguishable from human behavior.

Authors

  • Brenden M Lake
    Center for Data Science, New York University, 726 Broadway, New York, NY 10003, USA. brenden@nyu.edu.
  • Ruslan Salakhutdinov
    Department of Computer Science and Department of Statistics, University of Toronto, 6 King's College Road, Toronto, ON M5S 3G4, Canada.
  • Joshua B Tenenbaum
    Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139, USA. gershman@fas.harvard.edu horvitz@microsoft.com jbt@mit.edu.