Iterative weighted least squares algorithms for neural networks classifiers

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs) (2015, background): “Therefore tanh was previously preferred over logistic functions LeCun et al. (1991; 1998). Recently “Leaky ReLUs” (LReLUs) have been proposed, which replace the negative part of the ReLU with a linear function Maas et al. (2013). Leaky ReLUs have been shown to be superior to ReLUs in terms of learni”
Riemannian metrics for neural networks I: feedforward networks (2013, methodology): “3): at each unit, they solve a least-square problem of matching input signals and desired backpropagated output, an interpretation proposed in Kurita (1994). The quasi-diagonal reductions of these algorithms are based on the observation that there is a distinction between weights wik and wjk coming ”