Feature extraction

TG Tyler Grear
CA Chris Avery
JP John Patterson
DJ Donald J. Jacobs
request Request a Protocol
ask Ask a question
Favorite

The efficacy of a perceptron is modeled using a rectified adaptive nonlinear unit (RANU). For the k-th mode, the RANU is given by

where the quality factors Qd(k) and Qi(k) govern the strength of rectification, and the functions rd and ri quantify relevance. A mode is more relevant as Sk deviates farther from the bifurcation reference, Sm. Relevance is modeled as a function of x, where x=|ln(Sk/Sm)|. A linear rectifier is recovered when rd(x)=ri(x)=x. The nonlinear functions used in SPLOC-RNN are shown in Fig. Fig.22g.

Do you have any questions about this protocol?

Post your question to gather feedback from the community. We will also invite the authors of this article to respond.

post Post a Question
0 Q&A