Geek Answers Handbook

An example of a neural network for classifying multidimensional functions into two sets

I am looking for a good example of the source code of a controlled neural network that takes more than two functions (unlike most XY examples) and classifies the data into two sets. From what I read, there may be a Support Vector Machine (SVM) solution

All the classification examples that I found are two-dimensional. Here are a few:

I am trying to distinguish rare events from a series of inputs that are usually stable. Features - key-value pairs, where a value can usually be sampled as a small number. The available training data for the first category is huge, but with several sets of exercises for the second category, if that matters.

Sample training set

Category A

[2, 1, 0, 1, 4, 3] -> A [1, 1, 2, 3, 3, 0] -> A [0, 0, 1, 3, 2, 0] -> A

Category B

 [0, 4, 4, 4, 4, 3] -> B

Classification Example

 [1, 3, 4, 4, 4, 0] -> ??? (probably B)

Trust rating, for example. “85% defined B”, it would be useful to distinguish between a threshold for a rare event.

Is a neural network the best solution and are there any .NET libraries with this built-in?

+6
c # classification neural-network
source share
2 answers

In fact, all these machine learning methods have their pros and cons. When using NN (single-layer perceptron), you need to think about whether you have enough training data. From a technical point of view, you should be able to cover all cells within dimensions in order to have a good result.

SVM, on the other hand, is trying to find a border separating your data points, so if you have spaces in areas that are not close to that border, that's fine.

There are 5-6 classifiers around +/- boosting, and, to be honest, it seems that most of the time the type of classifier is selected subjectively. On the other hand, some people use several classifiers and compare the result.

In OpenCV, it’s so easy to connect another classifier so that you are on the right track for it. I used OpenCV in C ++ with NN classifiers for my project, and the result was very good:

http://www.springerlink.com/content/j0615767m36m0614/

+2
source share

SVM is n-dimensional - it's just that EXAMPLES are usually 2D, because once you get more than 3, the solution no longer fits into the 2D illustration.

It has only two output classes (usually Good and Bad), but it has as many features as you need. This is why the line separating your two SVM classes is called a “hyperplane” because it exists in multidimensional space — one dimension for each element.

-one
source share

More articles:


All Articles