Extended Data Table 3 Curriculum results for graph traversal

From: Hybrid computing using a neural network with dynamic external memory

  1. Parentheses represent ranges: (lower bound, upper bound). ‘Test’ is the accuracy (mean ± s.d.) on the test problem (here, the London Underground map) after the completion of each intermediate lesson. Evaluation of lesson completion occurs after every group of 100 batches has been processed on the main worker thread. The completion threshold is met if 90% of modal samples (most likely output of the network) are correct. ‘Final’ (mean ± s.d.) is the accuracy on the final lesson of the curriculum after the completion of each intermediate lesson.