25,161
edits
Changes
no edit summary
Finding the optimal point values of the weight vector θ,
<span style="font-size:150%;">''Θ '' =(''Θ''<span style="vertical-align: sub;">P</span>, ''Θ''<span style="vertical-align: sub;">N</span>, ''Θ''<span style="vertical-align: sub;">B</span>, ''Θ''<span style="vertical-align: sub;">R</span>, ''Θ''<span style="vertical-align: sub;">Q</span>)</span>
requires [https://en.wikipedia.org/wiki/Cross_entropy#Cross-entropy_minimization minimizing] the [https://en.wikipedia.org/wiki/Cross_entropy cross-entropy] [https://en.wikipedia.org/wiki/Cross_entropy#Cross-entropy_error_function_and_logistic_regression cost function] for the [https://en.wikipedia.org/wiki/Logistic_regression logistic regression] <ref>"Using [https://en.wikipedia.org/wiki/Cross_entropy#Cross-entropy_error_function_and_logistic_regression cross-entropy error function] instead of [https://en.wikipedia.org/wiki/Mean_squared_error sum of squares] leads to faster training and improved generalization", from [https://en.wikipedia.org/wiki/Sargur_Srihari Sargur Srihari], [http://www.cedar.buffalo.edu/~srihari/CSE574/Chap5/Chap5.2-Training.pdf Neural Network Training] (pdf)</ref>:
