Linear Self-attention Approximation via Trainable Feedforward Kernel.
Uladzislau YorshAlexander KovalenkoPublished in: ICANN (3) (2022)
Keyphrases
- feed forward
- neural network
- back propagation
- artificial neural networks
- recurrent neural networks
- closed form
- neural nets
- polynomial kernels
- hidden layer
- visual cortex
- kernel function
- biologically plausible
- artificial neural
- neural architecture
- kernel methods
- error tolerance
- activation function
- support vector
- linear functions
- recurrent networks
- approximation algorithms
- expert systems
- error back propagation