Linear layers in neural networks (NNs) trained by gradient descent can be expressed as a key-value memory system which stores all training datapoints and the initial weights, and produces outputs using unnormalised dot attention over the entire training experience. While this has been technically known since the 1960s, no prior work has effectively studied the operations of NNs in such a form, presumably due to prohibitive time and space complexities and impractical model sizes, all of them growing linearly with the number of training patterns which may get very large. However, this dual formulation offers a possibility of directly visualising how an NN makes use of training patterns at test time, by examining the corresponding attention weights. We conduct experiments on small scale supervised image classification tasks in single-task, multi-task, and continual learning settings, as well as language modelling, and discuss potentials and limits of this view for better understanding and interpreting how NNs exploit training patterns. Our code is public.
|Original language||English (US)|
|Title of host publication||ICML 2022 : 39th International Conference on Machine Learning|
|State||Published - Jun 17 2022|
Bibliographical noteKAUST Repository Item: Exported on 2022-12-21
Acknowledgements: This research was partially funded by ERC Advanced grant 742870, project AlgoRNN, and by Swiss National Science Foundation grant 200021 192356, project NEUSYM. We are thankful for hardware donations from NVIDIA & IBM. The resources used here were partially provided by Swiss National Supercomputing Centre (CSCS) project s1023.