Page - 97 - in Document Image Processing

J. Imaging 2018,4, 41 optimizeroutperformedtheSGDandAdagradoptimizers. TheAdaDelta,AdaMax,andRMSProp optimizersoutperformedtheAdamoptimizer. Figure5showstheperformanceof individualoptimizer. Table 5. In this table,we report the results in termofmaximum,minimum,mean, and standard deviation recognition accuracy obtained withNA-6 on ISIDCHARwhen the system trained for 50epochswith thedifferentoptimizers. Thebest scoresare inbold. RecognitionAccuracy DifferentOptimizers SGD Adagrad Adam AdaDelta AdaMax RMSProp Maximum 0.931 0.9364 0.9558 0.9565 0.9579 0.9602 Minimum 0.6933 0.7703 0.7585 0.7605 0.7851 0.8067 Mean 0.9168 0.9280 0.9411 0.9457 0.9448 0.9463 Std.Deviation 0.0365 0.0252 0.0274 0.0311 0.0256 0.0252 6*' (SRFK $GDJUDG (SRFK $GDP (SRFK $GD'HOWD (SRFK $GD0D[ (SRFK 5063URS (SRFK (a) (b) (c) (d) (e) (f) Figure5. In thisfigure,wedrawtherecognitionaccuracyobtainedwithNA-6networkarchitecture at different optimizers (a) SGD; (b)Adagrad; (c)Adam; (d)AdaDelta; (e)AdaMax; (f) RMSProp; onthe ISIDCHARdatabaseateachepoch. 97