Extracting and Utilizing Abstract, Structured Representations for Analogy
- Steven Frankland, Princeton Neuroscience Institute, Princeton University, Princeton, New Jersey, United States
- Taylor Webb, Princeton University, Princeton, New Jersey, United States
- Alexander Petrov, Department of Psychology, Ohio State University, Columbus, Ohio, United States
- Randall O'Reilly, Dept of Psych and Neuro, University of Colorado Boulder, Boulder, Colorado, United States
- Jonathan Cohen, Princeton Neuroscience Institute, Princeton University, Princeton, New Jersey, United States
AbstractHuman analogical ability involves the re-use of abstract, structured representations within and across domains. Here, we present a generative neural network that completes analogies in a 1D metric space, without explicit training on analogy. Our model integrates two key ideas. First, it operates over representations inspired by properties of the mammalian Entorhinal Cortex (EC), believed to extract low-dimensional representations of the environment from the transition probabilities between states. Second, we show that a neural network equipped with a simple predictive objective and highly general inductive biases can learn to utilize these EC-like codes to compute explicit, abstract relations between pairs of objects. The relational representations learned by the model can then be used to complete analogies involving the signed distance between novel input pairs (1:3 :: 5:? (7)), and extrapolate beyond the network's training domain. As a proof of principle, we extend the architecture to more richly structured tree representations.
Return to previous page