Metatrader Experts based on new Neural DLL
Neural MT4 EA Introduction
We are glad to bring to your attention new Neural Network Tools. This tool is based on the neuroX library. It is possible to realize neural network1 of practically any topology on the basis of this library. What are the advantages of Metatrader Expert Advisors, based on neural networks, over usual Metatrader Expert Advisors?Usual mt4 EA are programmed on the basis of the algorithm, which is based on certain trading strategy. Metatrader EA, which are based on neural networks, are not programmed in the full sense of this word: they are trained. Possibility of training is one of the main advantages of neural advisors over usual advisors. Training is a finding of links between neurons. In the course of training such Metatrader Expert Advisor is capable to reveal difficult dependences, and also to execute generalizations, and in case of successful training Expert Advisor can find true result on the basis of data, which were absent in the training sample.
If neural metatrader EA is trained correctly, it can predict future behavior of price on the basis of several previous values (patterns2) or on the basis of factors existing at the present moment.
Such Metatrader Expert Advisor is capable to allocate and generalize latent dependences; usual ЕА is incapable of doing it.
Accurate prediction is possible only when the previous changes somehow really predetermine the future. Therefore it is necessary to note that the neural network should be retrained after a certain operating period for more exact prediction.
Use of Kohonen maps3 in the network allows to examine the sample (price or other additional data) for many times. Each examination of sample is called “a training epoch”. Initial data (historical data, quotations) are divided into two parts during training: the training sample and the testing sample (as a matter of fact, the testing sample replaces forward test for such Metatrader Expert Advisor).
So, all our Neural Metatrader Expert Advisors consist or will consist of two basic nodes: neural net, organized by means of NeuralX.dll, and the advisor, which forms a signal for training.
Important! The trained network is saved in the directory Metatrader\tester\files\ and has the filename extension .nnb
All buyers of advisors, based on NeuroX.dll, can exchange the trained networks and discuss the work of advisors on our forum 24fx.us. We will also train and publish networks on the forum as far as possible.
MT4 NeuroX.dll Description of parameters (This information is given for the developers.)
There are several components in the neuroX.dll library for MetaTrader4, which are created in MQL4 language and used for realization of expert systems of analysis and prediction of the financial market.
The agreement on call of exported functions for scripting programs:
int NewNeuro( int NeuroType ); // Creation of object NeuroID >= 0 or -1 error
int NeuroAgent( string AgentName ); // Creation of AgentID >= 0 or -1 error
int NeuroLoad( string FileName ); // Creation and load of file NeuroID or -1
string GetProperty( int ObjectID, string Property ); // Reading parameters
bool SetProperty( int ObjectID, string Property, string Value ); // Record
bool NeuroSave( int NeuroID, string FileName ); // Saving data in file
bool NeuroInit( int NeuroID, double& Layers[], int Count ); // Initialization
// Input in network NeuroID data vector Data, OutCnt output и InCnt input
bool NeuroPattern( int NeuroID, double& Data[], int OutCnt, int InCnt );
// Network training using input data | Data, OutCnt output и InCnt input
bool NeuroTrain( int NeuroID, double& Data[], int OutCnt, int InCnt, int Num );
// Calculation in network NeuroID - return in Data, OutCnt outputs on InCnt inputs
bool NeuroCompute( int NeuroID, double& Data
Types of neural components - parameter NeuroType for function NewNeuro()
#define NT_SOM 1 // Self Org.Map Kohhonen.
#define NT_BPR 2 // Back Propagation
#define NT_HOP 3 // Net Hopfild
#define NT_LPF 4 // Low Pass Filter
#define NT_KLF 5 // Kalman Filter
#define NT_GEN 6 // Genetic Algorithm
#define NT_WNN 7 // Wizard Neural Network
Protection of Neural MT4 EA
Sale of advisors with open mql4 code was a policy of our company for several years. Lately we incur losses because of resale of our advisors through fraudulent web sites, exchange of advisors and their use in the championships under other names.In order to protect our intellectual property we were forced to protect our advisors by server-based guard. So, you can use the advisor only on one computer, but, at the same time, we do not limit the number of brokers or accounts, i.e. we do not limit the number of metatraders running on one computer (PC, VPS or Dedicated server). For registration of the bought advisor you just need to enter yours e-mail address in a registration window, which opens at the first start of the advisor.
In case of changes of hardware on your computer we will update information about new hardware and re-register you.
MT4 Expert Advisor "NeuralScalpingEA"
Now let’s speak about new NeuralScalping MT4 expert advisor in detail.
The expert advisor consists of the unique scalping expert advisor, which is based on in-depth analysis of the sample by means of overlap of multilayered flat channels that not only gives a chance to identify the price maximum and minimum on a certain time interval, but also allows to estimate the market tendency, i.e. if the market dies away or vice versa. Neural part also analyzes sample in the same time interval and takes a decision to open order or not to open it on the basis of the accumulated knowledge. Options also allow to open order at negative signal of the neural networks, but with changed StopLoss and TakeProfit and smaller lot. It allows to protect the deposit in a better way.
The expert advisor was created for work with currency pairs having low spread, for example, EUR/USD. In order to prevent the pernicious influence of the expert advisor on the market, the number of copies in sale is limited. MT4 EA NeuralScalping BackTest
Neural Forex knowledge base
MT4 Expert advisor Neural Training explanation Video1. Artificial neural networks — mathematical models and also their program or hardware realizations, created using principle of organization and functioning of biological neural networks —networks of nerve cells of a living organism. This concept was developed at studying of the processes proceeding in a brain at thinking, and at the attempt to simulate these processes.
2. Patterns – characteristic parts of time charts of prices, which are examined during the technical analysis.
3. Such networks represent the competitive neural network, working on visualization and clustering. It is a method of projecting the multidimensional space in space with lower dimension (two-dimensional); it is also used to solve problems of modeling and prediction.
Comments