Abstract
The success of deep learning, a brain-inspired form of AI, has sparked interest in understanding how the brain could similarly learn across multiple layers of neurons. However, the majority of biologically-plausible learning algorithms have not yet reached the performance of backpropagation (BP), nor are they built on strong theoretical foundations. Here, we analyze target propagation (TP), a popular but not yet fully understood alternative to BP, from the standpoint of mathematical optimization. Our theory shows that TP is closely related to Gauss-Newton optimization and thus substantially differs from BP. Furthermore, our analysis reveals a fundamental limitation of difference target propagation (DTP), a well-known variant of TP, in the realistic scenario of non-invertible neural networks. We provide a first solution to this problem through a novel reconstruction loss that improves feedback weight training, while simultaneously introducing architectural flexibility by allowing for direct feedback connections from the output to each hidden layer. Our theory is corroborated by experimental results that show significant improvements in performance and in the alignment of forward weight updates with loss gradients, compared to DTP. Show more
Permanent link
https://doi.org/10.3929/ethz-b-000465852Publication status
publishedExternal links
Book title
Advances in Neural Information Processing Systems 33Pages / Article No.
Publisher
CurranEvent
Organisational unit
09479 - Grewe, Benjamin F. / Grewe, Benjamin F.
Funding
173721 - Temporal Information Integration in Neural Networks (SNF)
189251 - Ultra compact miniaturized microscopes to image meso-scale brain activity (SNF)
186027 - Probabilistic learning in deep cortical networks (SNF)
Notes
Due to the Coronavirus (COVID-19) the conference was conducted virtually.More
Show all metadata
ETH Bibliography
yes
Altmetrics