Codifying temporal characteristics of Jewett components to improve Jewett transform

Jewett Transform is not yet, it is being. First ideas on this metaphor are from 1980 while monitoring cerebral function. It was conceived in contrast with Fourier Transform. Its application is limited to Auditory Brain Stem Responses. It uses a non-orthogonal physiologically rooted basis. Non-orthogonal basis has limited power in front of orthogonal basis: no analytical method exists to evaluate the corresponding transforms and numerical methods are required. In previous works, numerical methods were replaced for by trained artificial neural networks. Jewett transform was applied to increase the training set. Being a physiologically inspired basis, it promises better understanding of analysis of these evoked responses. It is envisioned that diverse new transforms, tailored to different problem specificity are to emerge. Considering the short temporal influence of Jewett components, it is stated that codifying temporal characteristics of Jewett components can be used to improve Jewett Transform. Previously used neural network was modified. Output vector codes are built up by grouping components instead of grouping parameters. This allows synaptic pruning in the artificial neural network. Only a fraction (0.49) of the previous network weights is used. Mean square error in fitting signal to model are acceptable (mean ε<0.3%, n= 600). Memorization is eliminated