Artificial neural networks. - page 7
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
I don't think it's all that hopeless really. The market has a tangible advantage over the visual cortex - the market lives in two-dimensional space, while people live in three-dimensional space. At any given time, the price can move either up or down. The results of the network in the market also have only two choices - right, wrong. This significant simplification of how the market works, compared to the visual cortex, is nonetheless encouraging.
If every movement was more than the spread + commission, yes. But in real life you don't have to guess the direction of the tick, but the price change of several spreads, in case of exchange trading it can be a hundred ticks.
If we move away from ticks and move to bars, a bar may be large, small, medium up and down or zero, how many measurements then. And the task is still the same, to predict a significant (compared to the spread) movement rather than a single price change.
I don't think it's really that hopeless. The market has a tangible advantage over the visual cortex - the market lives in two dimensions and people live in three dimensions. At any given time, the price can move either up or down. The results of the network in the market also have only two choices - right, wrong. Such a significant simplification of how the market works, compared to the visual cortex, is nonetheless encouraging.
I agree. There is no need to try to create an artificial brain that is the same as a living one. We have a living one ourselves. But the living brain has significant limitations - slow processing of sequential data. That is what we cannot do quickly (and also without the influence of emotions and other "worldly factors") - that is what we should try to simulate.
I know the words, it seems to be written in I.Toshchakov's book, but this is theory, in practice "these are every moments" end after entering the market, the time to hold a position or cancel a forecast is very important, the TS with well chosen historical take and stop ratios work poorly in the future, because the market dynamics/volatility is constantly changing
Although art for art's sake, also has the right to life.
I see, but it's still hundreds of thousands of orders of magnitude less than the visual cortex has to process in real life. But at the same time, we don't have those 140 million neurons in one layer.
Take chess as an example. Because it does not have an intellect, it just has algorithms for winning a very large number of known positions. That is positions in which there is a clear algorithm that leads to a victory. So, a machine wins a game to one of these positions, not with its intellect, but with its computing power and data base. Why can't this be done with NS in the marketplace?
To me, NS is the crowning glory of ideology of fitting in with history. Which is highly detrimental to sustainability. After all, the more pronounced extremum, the less chances (in the real world) that the process will revolve around it.
By optimization, I mean deleting unprofitable trades or adding profitable ones (which is much rarer) within the trading idea. If after optimization most of the entry/exit points (trades) have changed, then you have made a fit to the story, the NS in this case simply selected from the entire list, the parameters that best make this fit.
papaklass:
Не думаю, что все так безнадежно на самом деле. На рынке есть ощутимое преимущество перед зрительной корой головного мозга - рынок живет в двухмерном пространстве, а люди в трехмерном. В каждой момент времени цена может двигаться либо вверх, либо вниз. Результаты работы сети на рынке тоже имеют всего два варианта - правильно, неправильно. Такое существенное упрощение работы рынка в сравнении с работой зрительной коры, все-таки вселяет надежды.
Yes, what could be simpler than making binary decisions on a one-dimensional series :)
I wonder how many iterations the NS has to perform to find such a system
The simplest and failsafe strategy is "Free Cheese from FedReserve". On the day before the Fed meeting, buy the index at the close, assuming the close is below the 20-day high. The next day (when Fed announces the results of the meeting), sell at the close. Each trade is 100K, no reinvestment. That's it.
But the living brain has a significant limitation - slow processing of sequential data.
Who established this? We can categorise an object after 50 milliseconds with 80% accuracy. That's 20 objects per second, at any buckgrnd. Many mammals do it even faster to avoid being eaten (evolution). Artificial nets do it in a few seconds, and on an empty buckgrind. The power of the brain is in its parallelism, which we will never be able to achieve by conventional means of computer technology. No one denies the usefulness of automated trading, but networks will not replace the trader's brain in the search for patterns in the market in the next 20-30 years. It takes a lot of neurons. Does anyone here think that a network with 10-20 neurons can replace the trader's brain? What a stupid creature this trader must be!
How many years is that 210 deals?
How many years is that 210 deals?