mactivate-package {mactivate}

m-activation

Description

Provides methods and classes for adding m-activation ("multiplicative activation") layers to MLR or multivariate logistic regression models. M-activation layers created in this library detect and add input interaction (polynomial) effects into a predictive model. M-activation can detect high-order interactions – a traditionally non-trivial challenge. Details concerning application, methodology, and relevant survey literature can be found in this library's vignette, "About."

Details

The DESCRIPTION file:

Package:	mactivate
Type:	Package
Title:	Multiplicative Activation
Version:	0.6.6
Date:	2021-08-02
Author:	Dave Zes
Maintainer:	Dave Zes <[e-mail]>
Description:	Provides methods and classes for adding m-activation ("multiplicative activation") layers to MLR or multivariate logistic regression models. M-activation layers created in this library detect and add input interaction (polynomial) effects into a predictive model. M-activation can detect high-order interactions -- a traditionally non-trivial challenge. Details concerning application, methodology, and relevant survey literature can be found in this library's vignette, "About."
License:	GPL (>=3)
Depends:	R (>= 3.5.0)

Index of help topics:

df_hospitals_ortho      Orthopedic Device Sales
f_control_mactivate     Set Fitting Hyperparameters
f_dmss_dW               Calculate Derivative of Cost Function wrt W
f_fit_gradient_01       Fit Multivariate Regression Model with
                        mactivate Using Gradient Descent
f_fit_gradient_logistic_01
                        Fit Logistic Multivariate Regression Model with
                        mactivate Using Gradient Descent
f_fit_hybrid_01         Fit Multivariate Regression Model with
                        mactivate Using Hybrid Method
f_logit_cost            Logistic Cost
f_mactivate             Map Activation Layer and Inputs to Polynomial
                        Model Inputs
mactivate-package       m-activation
predict.mactivate_fit_gradient_01
                        Predict from Fitted Gradient Model
predict.mactivate_fit_gradient_logistic_01
                        Predict from Fitted Gradient Logistic Model
predict.mactivate_fit_hybrid_01
                        Predict from Fitted Hybrid Model

Please make sure to read Details in f_dmss_dW help page before using this library. This package allows the user to extend the usual multivariate regression solution by adding (parallel) multiplicative “activation layers.” These activation layers can be very useful for identifying input interactions, and, if the user wishes, transparently test the appropriateness of input transformations. Three functions are provided for fitting data, f_fit_hybrid_01 and f_fit_gradient_01 for a numeric response (usual MLR), and f_fit_gradient_logistic_01 for a binary response (multivariate logistic regresssion). The user is encouraged to consult the “About” vignette as well as the examples available in the respective functions' documentation for details about m-activation and practical examples of implementation.

Author(s)

Dave Zes

Maintainer: Dave Zes <[e-mail]>

Examples

## please see docs for individual functions.