Learning to Reason with Third-Order Tensor Products [article]

Imanol Schlag, Jürgen Schmidhuber
2019 arXiv   pre-print
We combine Recurrent Neural Networks with Tensor Product Representations to learn combinatorial representations of sequential data. This improves symbolic interpretation and systematic generalisation. Our architecture is trained end-to-end through gradient descent on a variety of simple natural language reasoning tasks, significantly outperforming the latest state-of-the-art models in single-task and all-tasks settings. We also augment a subset of the data such that training and test data
more » ... t large systematic differences and show that our approach generalises better than the previous state-of-the-art.
arXiv:1811.12143v2 fatcat:f77xc6qvqze3zn2dalp45ta7oq