Whether to use the activation function in the output layer or not OAF parameter value depends on the nature of outputs. The main disadvantage of gradient-based optimization methods is that they often find a local minimum. Select chart and Timeframe where you want to test your indicator. The output of the network is the predicted relative change of the next price.
If you were able to make a profitable trading system based on these codes, please share your idea with me by sending email directly to vlad yahoo. It consists of several layers: Here is an example of FFNN with one input layer, one output layer and two hidden layers: The data is processed by neurons in two steps, correspondingly shown within the circle by a summation sign and a step sign: All inputs are multiplied by the associated weights and summed The resulting sums are processed by the neuron's activation function , whose output is the neuron output.
This method is described here http: Examples of using the NN library: According to the Cybenko Theorem , a network with one hidden layer is capable of approximating any continuous, multivariate function to any desired degree of accuracy; a network with two hidden layers is capable of approximating any discontinuous, multivariate function: The optimum number of neurons in the hidden layer can be found through trial and error.
The following "rules of thumb" can be found in the literature: Keep track of the training error, reported by the indicator in the experts window of metatrader. For generalization, the number of training sets ntr should be chosen times the total number of the weights in the network. Therefore, the number of training sets ntr should be at least The concept of generalization and memorization over-fitting is explained on the graph below. The input data to the network should be transformed to stationary.
Forex prices are not stationary. It is also recommended to normalize the inputs to The indicator is universal, but it is better to use at higher timeframes. The indicator is presented in two forms: Please note, for proper operation of the indicator must be installed BPNN.
When switching timeframes indicator need to restart on graph, apparently due to the features of "neural networks". It is more convenient to do it through a template installation that I have prepared for you in the archive.
Hi, thank you for the indicator, having troubles with the dll file. Hi guys, indicator works, at least I was able to have prediction and red line.