Intel Neuron Bot

Intel Neuron Bot is an expert system based on a new architecture neural network. For the expert to work, a trained neural network file is required. Some files are ready on the expert's discussion tab. Consultations on the formation of a neural network file can be obtained in private messages or in the product discussion. The bot goes through the entire history on all currency pairs (provided that the trained neural network is loaded). The optimal working timeframe is H1.

The architecture of the neural network.

The neural network is based on a specially developed architecture (T-INN) Target-IntelNeuroNet. This architecture has advanced learning capabilities on large data sets. The T-INN architecture is distinguished by the fact that it can efficiently train on an unlimited data set while retaining the extrapolation properties of the data. With an increase in the data set, the training time increases, but it is incomparably less than with classical architectures (such as RBF, PNN, GRNN, MADALINE, MLP). And most importantly, with a set of input data of an unlimited size, the neural network will find acceptable solutions to the problem! What is absolutely impossible to achieve using classical neural networks in principle.

The basis of this architecture is a special structured approach to learning. All incoming training data is set simultaneously in the form of a large two-dimensional matrix, the neural network independently normalizes and structures this data into clusters, while a certain number of clusters are formed that combine data with similar patterns in certain characteristics, as a result of projecting a multidimensional space into a space from a two-dimensional reality. Thus, an array of matrices divided by certain classes of patterns is formed, the signs of each individual cluster can be in completely different parts of the base table and can be completely incoherent, the chronology is not saved in any way, while some data can be discarded outside the clusters and not used at all. To implement this part of the architecture, self-organizing Kohonen maps are used - a neural network with unsupervised learning. The next step uses a neural network based on a multilayer perceptron with two internal layers that work for each individual cluster. The cluster limits the training data in terms of volume so as to achieve the optimal performance of each individual section of the neural network.

Usage practice.

In practice, such a neural network can adapt to the entire history in 14 hours, for example, EURUSD with an accuracy of mse=0.0005. Approximate deviation on the training interval of the extrapolated function is 4%. For the forward period, this indicator reaches 20% deviation. Which is a very good result.

By default, the system works based on the direct outputs of the neural network, you can set the level of reaction to the neural network signal using the "Level Signal" field, and you can also set various EA parameters to level the chart, namely:

• ManyManagement
• stop loss
• take profit
• trailing start
• trailing stop
• TotalEquityLoss
• TotalEquityProfit
• AutoCloseArbitrary
• AutoCloseOnlyProfit
• AutoCloseOrder
• LimitOrders
• GridStep
• TypeSeria
• SeriaTrendOn
• SeriaAntyTrendOn

The Expert Advisor has the ability to create its own logs in a file and flexible configuration of comments within the metatrader. The Expert Advisor also has quite a lot of other settings in its arsenal that are necessary for normal operation on the market.