t201 opened a new pull request #961:
URL: https://github.com/apache/systemml/pull/961
This Documentation is for Multinomial Logistic Regression using Trust Region
method.
# Introduction
The DML (Declarative Machine Learning) language has built-in functions which
enable access to both low- and high-level function
## `Multinomial Logistic Regression` -Function
The Multinomial Logistic Regression Model is is used to link and provide an
estimate of category label probabilities ( y, having multiple rows single
column ) which are a response to the numerical vector of explanatory (feature)
variables (x, having multiple rows and multiple columns) . The number of links
can be 3 or greater unlike binomial regression.
### Usage
```
mulLogisticRegression = function( Matrix X, Matrix Y, intercept_scaling ,
max_iter , tolerance ,
max1 , maxi2 , verbose )
return(Matrix A)
### Arguments
| Name | Type | Default | Description |
| :---- | :------------- | -------- |
:------------------------------- |
| X | Matrix[Double] | -- | Matrix of numerical vector
of explanatory variables|
| Y | Matrix[Double] | -- | Matrix of a categorical
response variable|
| intercept_scaling| Int | 0 | intercept for shifting and
rescaling X columns|
| reg_para | Double | 0 | regularization parameter|
| tolerance | Double | 0.00001 | tolerance ("epsilon")|
| max1 | Int | 100 | max. no. of outer newton
interations|
| max2 | Int | 0 | max. no. of inner
(conjugate gradient) iterations|
### Returns
| Type | Description |
| :------------- | :---------- |
| Matrix[Double] | probability regression as output |
### Example
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
