Optimization of Deep Neural Network for Recognition with Human Iris Biometric Measure

DADCNet: Dual attention densely connected network for more accurate real iris region segmentation

“…The ultimate purpose of the research on iris segmentation algorithm is to improve the accuracy of the subsequent iris recognition process. Thus, some researchers did have not completed iris segmentation process, but directly recognize the original iris image after preprocessed, such as [19] and [20]. Preprocess can reduce the interference information of original iris image and improve the accuracy of subsequent iris recognition.…”

A novel method for structure selection of the Recurrent Random Neural Network using multiobjective optimisation

“…On the other hand, it is possible to find recent works where optimisation tools have been efficiently applied. For instance, the authors in [30] have used genetic algorithm to optimise the results in a deep neural network, where the number of neurons are the input parameters for the optimisation tool. Another example of optimised structure can be seen in [31], where structural parameters, such as the number of neurons are automatically selected to minimise the prediction error criterion according to Akaike's information criterion.…”