The recurrent neural networks, are much more effective and accurate, because their learning is correlated with each other, that is (it is a bit difficult to explain, we must abstract, and understand), the networks can consist of several groups of smaller networks , which send what they learn to the main portion of the network, and in turn the network contacts them, so it turns out that it indicates to these smaller portions if the information sent to the central is correct or not, all this during learning This is so true, that I taught in the new version of my Ea, from 2003 to 2018, and then I made a future bactext and won consistently until today. The network had no idea what would happen a year and a half later, it's a long time with the changes that happen every day in a lot. But it is the beginning of a development, which is in its second stage (look for Holy Grail Deep Neural Network Recurrent), you will find the first version today, I have to finish some security settings, to upload the web. Nothing I wrote here is 'lie.'
Thanks for reading me.