"New Neural" is an Open Source neural network engine project for the MetaTrader 5 platform. - page 64
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
Suggest another implementation, I'm not against it, lay out the idea of implementation, we'll discuss, compare, decide what's better.
Varum niimand arbaite?
Misha. You want some neurons? I have them.
You must have it made for yourself (the programmer). "You need to put the flag here, take it off here, I put the herring there, don't touch it yet.
And I need "for the dummies.
Write into csv text file all data in one column. Depending on architecture will be rules to read/write to that column. That's what I do now when training meshes - I store all network settings and weights in one column, it's equivalent to a chromosome representation and easy to read/write to GA.
So you change the architecture, change the rules?
And if the architecture is not created by a programmer, but a user, who will write a new rule?
Yes, and then for every sneaking is not enough, write for each architecture a new rule to load it (imho) burdensome.
Is it possible to have a variant of the algorithm that creates a loading rule depending on the architecture?
So you change the architecture, change the rules?
And if the architecture is not created by a programmer, but a user, who will write a new rule?
And you can't thank me for every sneaking, writing new loading rules for every architecture is (imho) very exhausting.
Is it possible to have a variant of the algorithm that creates a loading rule depending on the architecture?
I look at this problem in a simpler way.
Let's abstract from neurons in general. We have a set of some blocks - black boxes. We need to make a column of settings for all the blocks, so:
We ask the blocks how many settings you have (settings, weights, or whatever) and they answer:
1-й -10
2-й -115
3-й - 214
....
Then we write everything in a row in one column, and we always know from which place what goes.
This way we can always find out which settings belong to which block, to do this we just need to ask the block about it - "and how many settings you have.
You just need to agree right away in what order to poll the blocks, that's all.
SZZ Even if individual blocks will change by themselves in the future, nothing will be broken from this, and everything will work as before - the sequence of blocks polling is not to change. But even if you change the sequence, nothing will break, you just need to rewrite the column by polling the blocks again.
...
So the length of this array depends not only on the number and type of neurons, but also on the number of connections for each neuron?
because each neuron will need to prescribe to whom it is connected.
It turns out that the length of this array depends not only on number and type of neurons, but also on how many connections each neuron has?
because each neuron will need to prescribe to whom it is connected.
Of course. The length depends on what the blocks respond to.
Although... You're probably talking about data format, describing the architecture? - Here I'm of the opinion that it should be a separate file with its own format for writing to it.
And I was talking about all the settings/weights.
You must have it made for you (a programmer). "Here you need to put the flag, here you need to take it off, here I put a herring, do not touch it yet.
No way. Normally I have it done. True through the dll and on 4.
I need "for the dummies."
Sure. The length depends on what the blocks answer.
Although... You're probably talking about the format of the data describing the architecture? - Here I'm of the opinion that it should be a separate file with its own format for writing to it.
And I was talking about all the settings/weights.
loading algorithm is a converter from storage format to gridcreation format + grid creation algorithm.
We have dependencies...
the creation algorithm depends on the gridcreation format [-] ...
The gridcreationformat has no direct dependence [~] on the storage format, the dependence is indirect in the form of choosing the best one.
AZ=FX ~> FS -> AS
No, I don't. It's all right with me. True through the dll and on 4.
Write how to do it "for dummies". And for what you need also write.Tex. Personally, I know almost nothing about networks. From what I know, I think they are of no use to the market.
I want to make sure that this is true or to make sure that it is not true at all.
At the stage of acquaintance and mastering, I do not want to touch market quotes. I want to try the network on a sine wave with something I have defined; I want to try it on an hourly temperature chart, gradually incorporating data on pressure, time of year, humidity, wind strength and time of day.
With 95% accuracy I will indicate the temperature of the next hour based on these data, knowing their correlation. Will the grid be able to forecast one hour ahead, or two hours ahead, or more?
What you're giving me is nothing for the mt 4. You know how this box works, I don't.