skip to main content


Search for: All records

Creators/Authors contains: "Aiken, A."

Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?

Some links on this page may take you to non-federal websites. Their policies may differ from this site.

  1. Graph Neural Networks (GNNs) are based on repeated aggregations of information from nodes’ neighbors in a graph. However, because nodes share many neighbors, a naive implementation leads to repeated and inefficient aggregations and represents significant computational overhead. Here we propose Hierarchically Aggregated computation Graphs (HAGs), a new GNN representation technique that explicitly avoids redundancy by managing intermediate aggregation results hierarchically and eliminates repeated computations and unnecessary data transfers in GNN training and inference. HAGs perform the same computations and give the same models/accuracy as traditional GNNs, but in a much shorter time due to optimized computations. To identify redundant computations, we introduce an accurate cost function and use a novel search algorithm to find optimized HAGs. Experiments show that the HAG representation significantly outperforms the standard GNN by increasing the end-to-end training throughput by up to 2.8× and reducing the aggregations and data transfers in GNN training by up to 6.3× and 5.6×, with only 0.1% memory overhead. Overall, our results represent an important advancement in speeding-up and scaling-up GNNs without any loss in model predictive performance. 
    more » « less
  2. The neuronal cytoskeleton performs incredible feats during nervous system development. Extension of neuronal processes, migration, and synapse formation rely on the proper regulation of microtubules. Mutations that disrupt the primary α‐tubulin expressed during brain development,TUBA1A, are associated with a spectrum of human brain malformations. One model posits thatTUBA1Amutations lead to a reduction in tubulin subunits available for microtubule polymerization, which represents a haploinsufficiency mechanism. We propose an alternative model for the majority of tubulinopathy mutations, in which the mutant tubulin polymerizes into the microtubule lattice to dominantly “poison” microtubule function. Nine distinct α‐tubulin and ten β‐tubulin genes have been identified in the human genome. These genes encode similar tubulin proteins, called isotypes. Multiple tubulin isotypes may partially compensate for heterozygous deletion of a tubulin gene, but may not overcome the disruption caused by missense mutations that dominantly alter microtubule function. Here, we describe disorders attributed to haploinsufficiency versus dominant negative mechanisms to demonstrate the hallmark features of each disorder. We summarize literature on mouse models that represent both knockout and point mutants in tubulin genes, with an emphasis on how these mutations might provide insight into the nature of tubulinopathy patient mutations. Finally, we present data from a panel ofTUBA1Atubulinopathy mutations generated in yeast α‐tubulin that demonstrate that α‐tubulin mutants can incorporate into the microtubule network and support viability of yeast growth. This perspective on tubulinopathy mutations draws on previous studies and additional data to provide a fresh perspective on howTUBA1Amutations disrupt neurodevelopment.

     
    more » « less