statistics, data mining and machine learning toolbox

The issue occurs when using non-US locale.

The bug is in:

// XWilkinson.getList()
list.add(Double.valueOf(String.format("%." + Math.abs(digits) + "f", i)));

Double.valueOf(..) allows only numbers with dot as fraction separator. In case of Russion or French locale String.format() returns '0,123'.

I suggest using NumberFormat.getInstance().parse(String.format("%.4f", 0.123445)).doubleValue() to get consistent reuslts. Or replace String.fromat(..) with Formatter() which isn't locale specific.

To reproduce set local to non-us, e.g. add VM options

-Duser.language=ru
-Duser.country=RU

and run any code which invokes this method, e.g.

// kotlin
XWilkinson.base10(XWilkinson.DEEFAULT_EPS).searchBounded(0.0, 1.0, 5).list

P.S: Perhaps the same issue is not only in this method

Bumps notebook from 5.7.8 to 6.0.1.

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

Plot markers ignore number format settings defined in user locale. E.g. for RU or FR locale I get "0.7" instead of "0,7".

To reproduce set the locale to non-us, e.g. add VM options -Duser.language=ru -Duser.country=RU

Added Ridge regression. It allows the user to train and predict using Ridge Regression for a given Regularization strength Although I've only changed a few files, i seemed to have committed all the files, sorry about that, I am new to github, hope you understand

So, currently, TTestTwoSamples only allow testing by using the two whole samples.

I would like to be able to use the TTestTwoSamples if I know the mean and deviation of every sample already.

This is actually available in ZTestTwoSamples but not in TTestTwoSamples

Hope you consider adding this feature. Thanks in advance :)

Hello, I am trying to implement your library for a project. Everything is working really good, so thanks for creating this. Anyways, I need to get a value from the F distribution table given the alpha, the numerator degrees of variance, and the denominator degrees of variance.

Something like: F_0.025,9,14

Where: α = 0.025 num = 9 den = 14

The output value here should be 3.2093 according to this table

I am not sure if this is already possible and I just couldn't find the right class and method to call. If that's the case, would you mind explaining me what to use to achieve this?

If this isn't available yet, could you please consider this as a feature request?

Thanks again for such an amazing work.

1. In NaiveBayes.java : duplicated code in coreFit() method delete => BuildSumLog() method

2. extract class : NaiveBayesData, NumericData, NominalData, BinaryData

1. Extract 3 methods in byRows() method (BoundFrame.java)

   • nameLengthComp()
   • nameValueComp()
   • columnExistsCheck()

2. Renaming 2 methods in BoundFrame.java

   • rowCount() to getrowCount()
   • varCount() to getvarCount()

It is useful that sometimes to be able to grow dynamically the size of the vector. similar with how the java.util.Array behaves. However, when one knows the appropriate initial allocation size, to be able to use that.

The AdoptOpenJDK has been discontinued since July 2021. When using Zulu you get all the latest updated (TCK Tested) builds for all versions of OpenJDK. Added the upcoming LTS release JDK 17.

Bumps notebook from 5.7.6 to 5.7.8.

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

Introduce class priors for classification. Currently class priors are either simulated, only for learning, through individual weights, or are implemented as a custom feature in some of the classifiers. Class priors should be a generic feature used in scoring methods. As such, we should implement that in a more generic way.

Currently the graphics package allows one to draw axis element on left (for vertical axis) and bottom (for horizontal axis). This is enough for simple plots, but further customization will allow one o draw multiple plots on the same grid layer cell and also nicer drawings in general.

The last big refactory for graphical components allows one to have multiple plots superimposed on the same grid layer cell, each plot with different pair of axes. Technically this is possible, it depends however on the ability to draw on left,right,top, bottom the axes, but we need also a nice API and a decent default behavior for simple situations.

A violin plot is similar to a kernel density estimator (actually it is a kernel density estimator), but displayed in parallel, like in boxplot, not superimposed like one can normally already is able to do (multiple kernel density estimator lines). The violin chart is like a boxplot but with more flexibility and carried information about the density.