This　paper　proposed　an　idea　and　presented　a　scheme　of　using　neural　network　to　implement　real　coded　genetic　algorithm　(GA).　The　arithmetical　crossover　is　generated　to　its＇　multiple　parent　version　and　the　random　mutation　operator　is　also　generated　to　its＇　multiple　mutation　point　version,　two　artificial　neuron　models　are　also　designed　for　the　two　genetic　operations.　We　validate　our　scheme　with　some　benchmark　functions.　The　significance　of　our　research　means　that　the　GA　can　be　implemented　with　hardware　and　the　inherent　parallelism　of　GA　can　be　explicitly　realized,　as　a　result,　the　real-time　performance　of　GA　is　remarkably　improved,　and　the　application　field　of　GA　is　widely　broadened.　©　2003　IEEE.