"New Neural" is an Open Source neural network engine project for the MetaTrader 5 platform. - page 76
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
Probably a silly question.
Is it possible to classify vectors whose dimensionality is not equal to N with the Kohonen map tuned to a vector of dimension N. Basically a person will classify a sphere with a circle, a square with a cube, a pyramid with a triangle. I hope the idea is clear.
No, the idea is not clear. A person reads video information with the same matrix of inputs. There are neither more nor less receptors in the eye.
You propose to give vectors of different dimensions, how can we expect the grid to respond adequately ???
Probably a silly question.
Is it possible to classify vectors whose dimensionality is not equal to N with the Kohonen map tuned to a vector of dimension N. Basically a person will classify a sphere with a circle, a square with a cube, a pyramid with a triangle into the same class. I hope the idea is clear.
If you look at the cone from below, it is a circle and from the side it is a triangle.
The cylinder can also be rotated.
If you look at the cone from below, it is a circle, and from the side it is a triangle
The cylinder can also be rotated.
I apologize, I am new to neurogames, so maybe these questions are stupid.
There is a set of vectors X1,X2 . X1={x1,x2,x3,} . And x1>> x3 , x2>>x3 . x1 and x2 of both vectors are approximately the same .Then it turns out that in space these two vectors will be near . Although x3 describes the most important characteristic. For our rams x1=period average, x2=period PSY, x3=dispersion, x4=trend component. Then in this case classification will be mainly based on x1 and x2 . How to avoid this situation or am I slowing down intensively again.
I apologize, I am new to neurogames, so maybe these questions are stupid.
There is a set of vectors X1,X2 . X1={x1,x2,x3,} . And x1>> x3 , x2>>x3 . x1 and x2 of both vectors are approximately the same .Then it turns out that in space these two vectors will be near . Although x3 describes the most important characteristic. For our rams x1=period average, x2=period PSY, x3=dispersion, x4=trend component. Then in this case classification will be mainly based on x1 and x2 . How to avoid this situation or am I slowing down intensely again.
Can you give a concrete example and set a specific problem ?
That way it's easier to understand the essence of the question
Can you give a concrete example and set a specific task?
It's easier to understand the question.
The task is to choose a vector that would divide the market into clusters: Trend Up, Trend Down, Trend East.
For example let's form an array of vectors according to our fantasy X{x1,x2,x3,x4,x5,x6,x7}
x1= period MA
x2= period of MA
x3=value of MA
x4=value of MA
x5= dispersion e.g.
х6= trend component of MA(N)-MA(N-1)
х7=number of intersections of MA and price
If we cluster the array of these vectors using the Kohonen map, we will see closely related vectors. It turns out that x1,x2,x4,x7 will have the greater influence on the Euclidean distances. Although characteristics x3,x5,x6 are no less if not more important. We can normalize all the x's in the interval -1...1, but I don't know how to do it. Or you can take market characteristics close to their values, in this case we get a comparison of flies with cutlets.
The task is to select a vector that would divide the market into clusters of Trend Up, Trend Down, and Trend Out.
For example we form an array of vectors according to our fantasy X{x1,x2,x3,x4,x5,x6,x7}
x1= period MA
x2= period of MA
x3=value of MA
x4=value of MA
x5=dispersion e.g.
х6= trend component of MA(N)-MA(N-1)
х7=number of intersections of MA and price
If we cluster the array of these vectors using the Kohonen map, we will see closely related vectors. It turns out that x1,x2,x4,x7 will have the greater influence on the Euclidean distances. Although characteristics x3,x5,x6 are no less if not more important. We can normalize all the x's in the interval -1...1, but I don't know how to do it. Or if we take market characteristics close to their values, then we get a comparison of flies with cutlets.
Horses and people get mixed up... The MA period, the MA value ...
Maybe try to use ready-made programs and their help first, to understand the problems?
Deductor, NS2
Horses and people are all mixed up. MA period, MA value
Maybe try to use ready-made programs and their help first, to understand the problems?
Deductor, NS2
I agree that this example is not quite right. Then another - cluster the elderly horses and young basketball players by height, weight, age. It is kind of clear that any new object will be assigned to its own cluster. Now weight=1/real weight. The situation when height and age are approximately the same, and weight is much less than height and age (comparing flies with cutlets). It turns out that weight in this case has almost no effect on the vector and the basketball player is indistinguishable from the horse.
The task is to select a vector that would divide the market into clusters of Trend Up, Trend Down, and Trend Out.
For example we form an array of vectors according to our fantasy X{x1,x2,x3,x4,x5,x6,x7}
x1= period MA
x2= period of MA
x3=value of MA
x4=value of MA
x5=dispersion e.g.
х6= trend component of MA(N)-MA(N-1)
х7=number of intersections of MA and price
If we cluster the array of these vectors using the Kohonen map, we will see closely related vectors. It turns out that x1,x2,x4,x7 will have the greater influence on the Euclidean distances. Although characteristics x3,x5,x6 are no less if not more important. We can normalize all the x's in the interval -1...1, but I don't know how to do it. Or we can take market characteristics that are close to each other, in this case we have a comparison of flies with cutlets.
Vectors will be close to :
X1{10,13,26,12,42,48,98} and
X2{11,12,27,14,43,46,88}, and vector X3 is in another cluster
X3{101,12,27,14,43,46,88}
although not the fact, I have shown how to divide the clusters by Hamming distance, how the parameters of vectors will look if they are divided by the principle of "Trend Over, Trend Down, Trend Out" only FF knows :)