Neural Networks - Genetic Algorithms - Boltzmann Machines - FOREX crystal balls?

[lumanauw]

Thanks. But I still don't understand that Wikipedia article. Is there a simple explenation for what it is?

Neural Network seems tobe a program that looks for patterns and memorize those patterns. How a Genetic Alogarithm differs from this?

Is there any example of Genetic Alogarithm on MetaTrader (EA)?

 

[CharlesSecGen]

Thanks. But I still don't understand that Wikipedia article. Is there a simple explenation for what it is?

Try Genetic Algorithms in Plain English.

Neural Network seems tobe a program that looks for patterns and memorize those patterns. How a Genetic Alogarithm differs from this?

Neural networks are completely different animals and generally not considered to be in the evolutionary computation realm. Think of genetic programming and genetic algorithms as a biological process applied to non-biological problems. Neural networks replicate the connection methods of the brain in a computer. What factors do you consider when contemplating an investment? Now move that thought process to a computer.

Humans are great at identifying patterns. Neural networks move that activity to a machine.
[Of course, experts in any of these matters are now choking on my rudimentary explanation.]

 

[lumanauw]

I've read the last url, it is more clear than the wikipedia one. Genetic Algorithms is selecting possible way outs plus with their mutations. Am I right here?

I think it is good for looking answers for history data.

But something tickles me. How effective is this headed to new untrained data set? Do you have a # of success rate of this method headed with new untrained data?

 

[lumanauw]

In standard MT4 backtesting / optimizing, the standard option is using "Genetic Algorithm" to get faster and to the point results. Is this the same Genetic Algorithm we are talking here?

 

[CharlesSecGen]

Genetic Algorithms is selecting possible way outs plus with their mutations. Am I right here?

The typical genetic algorithm identifies the most fit member from the population at each generation. At some point the program stops the search and presents the user with the final selection.

I think it is good for looking answers for history data.

That's all you have to work with.

How effective is this headed to new untrained data set? Do you have a # of success rate of this method headed with new untrained data?

It depends upon the particular implementation of the algorithm. Evolutionary computation methods are simply tools that allow you to forego brute force methods of finding the solution. You could have a computer step through each possible permutation but you don't have that much time. Genetic programming software finds a reasonable solution in a fraction of the time.

In standard MT4 backtesting / optimizing, the standard option is using "Genetic Algorithm" to get faster and to the point results. Is this the same Genetic Algorithm we are talking here?

Probably. You'll have to ask them.

 

[lumanauw]

Hi,

Thanks for the explenation. I found this somewhere on the net. The title said it is somehow related to Genetic Algorithm. Unfortunately, I don't understand the codes, the explentions are in Russian language, not catched by MetaEditor.

Anybody can re-write (in English) the metaeditor comments (//........), what is each section of the code doing?

Is it really a Genetic Algorithm MT4 EA?

 

[C Little]

Here's a a few tips when using neural networks.

1) standardise the inputs by using the natural log of lagged returns (eg take the log of the difference in closing prices over x bars)
2) look for correlations between a target variable of AT LEAST 50 bars in the future, and inputs using lagged returns covering the past 50-250 bars
3) use intermarkets as the primary source of inputs
4) keep the number of inputs below 7-8 (as few as possible)
5) do not use more than 1 hidden layer
6) build several networks and use emsembles
7) do not try and predict price as the target variable - rather predict the log return of price over the next x days, or something like the rsi.
8) stay away from predicting moving averages - they are easy to predict and utterly useless in real trading (one particular software vendor uses predictions of moving averages and i know from bitter experience its a pointless exercise)
9) before building a network do seperate analysis on correlations

I have reserached neural nets for some time and have had some success - particularly with forex. (e.g over 90% directional accuracy at predicting direction over the next 50 bars on out of sample and live trading). Its not easy, and every market behaves differently, so no holy grail.

happy trading.

Could you provide a list of software/reading material you used while researching neural nets. Many Thanks.

 

[neuralforexsystem]

On my website I show the performance from my own neural network forecasting the Eurusd on daily basis. The key is with which inputs/variables you feed the NN. My experience says I have never achieved a good performance using indicators like MACD, RSI, etc.... to feed my NN. Better look at the price increments for example in a short period.

Maybe this link can be useful ( I haven't tested their soft ) :

An Introduction to Neural Networks, including Neural Network FOREX Trading Tutorial, Neural Network Software, Neural Network Example Code for FOREX Trading, as well as Neural Networks Programming Guide