This article defines the concepts used for the comparative case study of liquidity risk measurement for the Swedish bond market, presented in the second piece of a given series. The concepts of neural networks, back-propagation and numerical optimisation algorithms are outlined. This is followed by an introduction to SNN, which applies a scaled exponential linear function to address the common issue of overfitting. Lastly, a more basic logistic regression approach is presented as an alternative method.
An Introduction To Self Normalizing Neural Networks
Download PDFAn Introduction To Self Normalizing Neural Networks
Download PDF