Improved Arabic Alphabet Characters Classification Using Convolutional Neural Networks (CNN)

Table 4

Experimental results on the Hijja and AHCD through different optimizers algorithms.

Algorithm	Accuracy (%)				Precision (%)				Recall (%)				F1 score (%)
	AHCD		Hijja		AHCD		Hijja		AHCD		Hijja		AHCD		Hijja
	Training	Testing	Training	Testing	Training	Testing	Training	Testing	Training	Testing	Training	Testing	Training	Testing	Training	Testing
Adam	99.76	96.66	97.60	88.46	99.76	96.75	97.40	87.98	99.76	96.67	97.60	88.46	99.76	96.67	97.60	88.47
SGD	99.44	95.68	94.16	87.85	99.45	95.84	94.28	87.93	99.45	95.68	94.16	87.85	99.45	95.70	94.17	87.88
RMSprop	99.72	96.22	96.55	87.98	99.73	96.30	94.28	87.86	99.72	96.22	94.16	87.98	99.72	96.23	94.17	87.95
AdaGrad	88.62	83.54	63.90	60.37	88.73	83.78	87.86	64.95	88.62	83.54	87.98	63.91	88.60	83.52	87.95	63.77
Nadam	99.77	96.73	97.76	88.57	99.77	96.80	97.57	87.86	99.77	96.73	97.76	87.98	99.77	96.72	97.76	87.95
Momentum	99.16	95.56	94.16	87.76	99.18	95.72	94.11	87.73	99.17	95.57	94.17	87.77	99.17	95.58	94.15	87.74
AdaMax	99.89	96.19	97.19	87.57	99.90	96.25	97.01	87.86	99.90	96.19	97.19	87.98	99.90	96.19	97.20	87.95