sergiofierens / ai4r Goto Github PK

@SergioFierens

I took part of your code that was ported to Crystal-lang for an ANN project I'm doing (MIT licensed), just wanted to make sure it's ok with you code here I gave you and the repo credits for the work.

Hope it's ok.

Support request -- sorry!

I'm using the classifier on a data set pulled from a database. Let's say it looks like this (details don't really matter)

labels = [quiz1_points, quiz2_points, quiz3_points, quiz4_points]
data = [
[10,3,0,22], # this is user_id = 15
[5,1,5,16], # this is user_id 55
etc..
]

The classifier works great to sort the users into, say 4 groups.

Is there any way to get the user_ids for each group without looping through and using eval([10,3,0,22]) to find that its in group 1?

Is that what a "class value" is for? Data sets have the last element as the "class value"..

Can I use this gem to help me to create an specialist system to dectec wich disease is acometting an animal? Is this the best option, or do you know another gem that is better for this purpose ?

README suggests there's API docs at www.ai4r.org/rdoc/index.html, but every URL at that domain redirects back to GitHub.

Hi there, I'd love to use this gem but the lack of a license means I'm not able to. Would you mind adding a license to this repository? http://choosealicense.com has a good overview of some possible options.

Thanks! 😄

Hi,

There are 3 failing tests on HEAD, is it expected? The following commands were used:

$ git clone git://github.com/SergioFierens/ai4r.git
Cloning into 'ai4r'...
remote: Counting objects: 1165, done.
remote: Compressing objects: 100% (509/509), done.
remote: Total 1165 (delta 715), reused 1067 (delta 622)
Receiving objects: 100% (1165/1165), 370.14 KiB | 487 KiB/s, done.
Resolving deltas: 100% (715/715), done.
$ cd ai4r/
$ rake test

# Running tests:

..............................................................FF.......................F........................

Finished tests in 0.405638s, 276.1083 tests/s, 1338.6320 assertions/s.

  1) Failure:
test_eval(HyperpipesTest) [/Users/carlosagarie/Projects/ai4r/test/classifiers/hyperpipes_test.rb:64]:
<"N"> expected but was
<"Y">.

  2) Failure:
test_get_rules(HyperpipesTest) [/Users/carlosagarie/Projects/ai4r/test/classifiers/hyperpipes_test.rb:80]:
<"N"> expected but was
<"Y">.

  3) Failure:
test_eval(MultilayerPerceptronTest) [/Users/carlosagarie/Projects/ai4r/test/classifiers/multilayer_perceptron_test.rb:54]:
<"Y"> expected but was
<"N">.

112 tests, 543 assertions, 3 failures, 0 errors, 0 skips
rake aborted!
Command failed with status (3)

When using numeric data types in data_set.rb, a range (min and max values) is provided for the domain. When naive_bayes performs counts on count_instances, value_index returns nil because Numerical values (not equal to the max or min) are not contained in @values.

What license is this code publisehd under?

Hi,
Is there any plan to add DBscan as cluster algorithm ? If no, would you accept PR of this method ?

I have a large data set containing urls and some values for each of the urls. Upon clustering, I can see a bunch of clusters along with the data_items, but is there a way to retrieve the original url for a given set of values?

Currently, i have to reverse lookup in the original data structure, and it is not efficient as multiple urls have exact same set of values.

Does AI4r have any way to let me detect the number of clusters present/inherent in a dataset (for use with a clustering algorithm), for example I would like to be able to say something like:

Best fit: 5 clusters, 99% suitable
Second best fit: 4 clusters, 88% suitable.

Currently I require a user to enter the number of clusters but this is just guess work then...

NameError: uninitialized constant Ai4r::Classifiers::NaiveBayes::DataSet
    from /usr/local/var/rbenv/versions/1.9.3-p392/lib/ruby/gems/1.9.1/gems/ai4r-1.12/lib/ai4r/classifiers/naive_bayes.rb:105:in `build'

There is little typo in Ai4r::Classifiers::NaiveBayes#build method causing NameError.
PR attached.

Please link/host/whatever the original site somewhere again. The domain listed everywhere didn't get renewed as it seams

The latest version of the ai4r gem on rubygems.org is 1.09, even though 1.10 has been out since 2010.

A quick search on rubygems.org shows that the gem name is now 'ai4ruby' (I'm guess that the author changed, so the previous gem 'ai4r' in inaccessible) at version 1.14.

Could the docs please be updated so that people (e.g., myself) do not mistakenly install the old gem with its known 1.9.x serialization issues?

Hello,

Is it possible to get same k-means clusters for every execution of
Ai4r::Clusterers::KMeans.build(set, number_clusters)
for a particular data set?

If you are familiar with R, this can be achieved by using set.seed(val). More info here.

Thanks!

ZeroR chooses the first class value it comes across, not the most frequently accessed one. The test suite is a bit flawed in that the first class value also happens to be the most frequently accessed one.

The core problem is that no frequency is ever calculate so every class value looks like it has a frequency of 1.

I've fixed the problem on this branch:
http://github.com/nirvdrum/ai4r/tree/fix_zeror

With this commit:
ca6a404

http://ai4r.rubyforge.org/ is down.