[keghn]

 Here are four: 
1). k nearest Neighbor: 
 

2). k means : 
main classification is in the middle of a cluster. 
 
 
 

3). Cascade Classification, Haar: 
http://docs.opencv.org/2.4/modules/objdetect/doc/cascade_classification.html   
 
4). Neural Networks: 
 

[keghn]

The Curse of Dimensionality in classification: 

http://www.visiondummy.com/2014/04/curse-dimensionality-affect-classification/

[yotamarker]

interesting read, at any rate it seems unlikely to me that nature uses something that complex.

[Korrelan]

@Keghn...I'm pleased he realised/ stated...

In fact, this depends on the amount of training data available, the complexity of the decision boundaries, and the type of classifier used.

If the theoretical infinite number of training samples would be available, the curse of dimensionality does not apply and we could simply use an infinite number of features to obtain perfect classification.

 

[keghn]

 I have giving it thought about making it work in  x, y, z dimensions. Such as they do with PCA and t-SNE mapping. 

 With every object having their features. Each feature and sub feature is a dimension. Such as a outline of a box. That has four corners and four edges which has eight dimensions. 
 The flower tutorial above was nice. it only have two dimensions. 

 But the way i am dealing with it, in this universe, is to pair all features and sub features with weight. The weight is A value 
multiplied to a sub feature. It is initially set to one.   

 Now for example, you have two objects  and each object is described by its sub features, such as 
corners and straight edges of length, and curved line of length. 
 Then to find the distance in n space is to select one object and start increasing or decreasing weight until it become the 
other object. 

 The amount of change it to the weight is "the DISTANCE". 

 With million and billion of objects in this world, it is more like of being a hyper space pilot and selecting a conduit to another
object. 
 And when you do move keep track of what you moving away from and what you are move to. There is no x, y, and z here it is
relative relation with each other objects. Like the mother nature does it.

 But there are trick to view it in flat space.
PCA: 
https://en.wikipedia.org/wiki/Principal_component_analysis 

http://ccwu.me/vsfm/ 


t-SNE: 

[yotamarker]

[yotamarker]

I want the alg that makes the optical mouse detect if it was moved left or right

[keghn]

 This total software problem. it would have to built out of something that would get the mouse position. That is if
you are using visual basic?:

Private Sub Form1_MouseClick(ByVal sender As Object, ByVal e As System.Windows.Forms.MouseEventArgs) Handles Me.MouseClick
    Dim MPx As Point = MousePosition()
    TextBox1.Text = MPx.ToString

End Sub

[yotamarker]

no I was talking about image motion direction recognition, I thought about using the mouse's  algorithm to solve it.
I found no proper walkthrough online so I figured it out myself, but the thing was that alg is only accurate as long as
the whole image moves so I modified it, I'LL upload some vids sooner than later, thanks.

[yotamarker]

1st time I used static dim 

[yotamarker]

[keghn]

AMD Threadripper—16 cores and 32 threads for $999–arrives in August:   

https://arstechnica.com/gadgets/2017/07/amd-threadripper-16-cores-and-32-threads-for-999-arrives-in-august/

[yotamarker]

the direction getter code itself is actually 10000 times faster then the image recognition

[yotamarker]

when I speed up the image recog some objects aren't fully detected like the circle is detected as two half circles

[keghn]

In the math are you using the length of curve and the chord length?
 If so then curve length is an arc of 70 degrees then it's mirror would be 190 degrees or other
wise of putting it 70 and -70 degree: 

https://www.ck12.org/trigonometry/Length-of-a-Chord/lesson/Length-of-a-Chord-TRIG/