Fits model to data

Description

Generic function to fit models (inherited from mlapiEstimation)

Usage

fit(x, model, y = NULL, ...)

## S3 method for class 'Matrix'
fit(x, model, y = NULL, ...)

## S3 method for class 'matrix'
fit(x, model, y = NULL, ...)

Arguments

Value

invisible(object$self())

Fit model to the data, then transforms data

Description

Generic function to fit transformers (inherits from mlapiTransformation)

Usage

fit_transform(x, model, y = NULL, ...)

## S3 method for class 'Matrix'
fit_transform(x, model, y = NULL, ...)

## S3 method for class 'matrix'
fit_transform(x, model, y = NULL, ...)

Arguments

Value

Transformed version of the x

Base abstract class for all decompositions

Description

Base class for all decompositions which are methods which can decompose matrix into 2 low-dimensional matrices x = f(A, B). (Think of this Latent Dirichlet Allocation, Non-negative Matrix Factorization, etc). It iherits from mlapiTransformation and additionally requires to implement components member.

Base class for all decompositions which are methods which can decompose matrix into 2 low-dimensional matrices x = f(A, B) incrementally. It iherits from mlapiDecomposition and additionally requires to implement partial_fit method which can learn components incrementally.

Usage

mlapiDecomposition

mlapiDecompositionOnline

Format

R6Class object.

Fields

components

features embeddings. So if matrix is decomposed in a form x = f(A, B) where X = n\*m, A = n\*k, B = k\*m them B = components

components

features embeddings. So if matrix is decomposed in a form x = f(A, B) where X = n\*m, A = n\*k, B = k\*m them B = components

Methods

$fit_transform(x, y = NULL, ...)

$transform(x, ...)

Performs transformation of the new data (after model was trained)

$fit_transform(x, y = NULL, ...)

$partial_fit(x, y = NULL, ...)

$transform(x, ...)

Performs transformation of the new data (after model was trained)

Arguments

x

A matrix like object, should inherit from Matrix or matrix. Allowed classes should be defined in child classes.

y

NULL. Optional taget variable. Usually this should be NULL. There few cases when it could be used.

...

additional parameters with default values

x

A matrix like object, should inherit from Matrix or matrix. Allowed classes should be defined in child classes.

y

NULL. Optional taget variable. Usually this should be NULL. There few cases when it could be used.

...

additional parameters with default values

Examples

TruncatedSVD = R6::R6Class(
  classname = "TruncatedSVD",
  inherit = mlapi::mlapiDecomposition,
  public = list(
    initialize = function(rank = 10) {
      private$rank = rank
      super$set_internal_matrix_formats(dense = "matrix", sparse = NULL)
    },
    fit_transform = function(x, ...) {
      x = super$check_convert_input(x)
      private$n_features = ncol(x)
      svd_fit = svd(x, nu = private$rank, nv = private$rank, ...)
      sing_values = svd_fit$d[seq_len(private$rank)]
      result = svd_fit$u %*% diag(x = sqrt(sing_values))
      private$components_ = t(svd_fit$v %*% diag(x = sqrt(sing_values)))
      rm(svd_fit)
      rownames(result) = rownames(x)
      colnames(private$components_) = colnames(x)
      private$fitted = TRUE
      invisible(result)
    },
    transform = function(x, ...) {
      if (private$fitted) {
        stopifnot(ncol(x) == ncol(private$components_))
        lhs = tcrossprod(private$components_)
        rhs = as.matrix(tcrossprod(private$components_, x))
        t(solve(lhs, rhs))
      }
      else
        stop("Fit the model first woth model$fit_transform()!")
    }
  ),
  private = list(
    rank = NULL,
    n_features = NULL,
    fitted = NULL
  )
)
set.seed(1)
model = TruncatedSVD$new(2)
x = matrix(sample(100 * 10, replace = TRUE), ncol = 10)
x_trunc = model$fit_transform(x)
dim(x_trunc)

x_trunc_2 = model$transform(x)
sum(x_trunc_2 - x_trunc)

#' check pipe-compatible S3 interface
x_trunc_2_s3 = transform(x, model)
identical(x_trunc_2, x_trunc_2_s3)

Base abstract class for all classification/regression models

Description

Base class for all estimators. Defines minimal set of members and methods(with signatires) which have to be implemented in child classes.

Usage

mlapiEstimation

Format

R6Class object.

Methods

$fit(x, y, ...)

$predict(x, ...)

Makes predictions on new data (after model was trained)

Arguments

x

A matrix like object, should inherit from Matrix or matrix. Allowed classes should be defined in child classes.

y

target - usually vector, but also can be a matrix like object. Allowed classes should be defined in child classes.

...

additional parameters with default values

Examples

SimpleLinearModel = R6::R6Class(
classname = "mlapiSimpleLinearModel",
inherit = mlapi::mlapiEstimation,
public = list(
  initialize = function(tol = 1e-7) {
    private$tol = tol
    super$set_internal_matrix_formats(dense = "matrix", sparse = NULL)
  },
  fit = function(x, y, ...) {
    x = super$check_convert_input(x)
   stopifnot(is.vector(y))
   stopifnot(is.numeric(y))
   stopifnot(nrow(x) == length(y))

   private$n_features = ncol(x)
   private$coefficients = .lm.fit(x, y, tol = private$tol)[["coefficients"]]
 },
 predict = function(x) {
   stopifnot(ncol(x) == private$n_features)
   x %*% matrix(private$coefficients, ncol = 1)
 }
),
private = list(
  tol = NULL,
  coefficients = NULL,
  n_features = NULL
))
set.seed(1)
model = SimpleLinearModel$new()
x = matrix(sample(100 * 10, replace = TRUE), ncol = 10)
y = sample(c(0, 1), 100, replace = TRUE)
model$fit(as.data.frame(x), y)
res1 = model$predict(x)
# check pipe-compatible S3 interface
res2 = predict(x, model)
identical(res1, res2)

Base abstract class for all classification/regression models which can be trained incremendally (online)

Description

Base class for all online estimators. This class inherits from mlapiEstimation and additionally requires to implement $partial_fit(x, y, ...) method. Idea is that user can pass x, y in chunks and model will be updated/refined incrementally.

Usage

mlapiEstimationOnline

Format

R6Class object.

Methods

$fit(x, y, ...)

$partial_fit(x, y, ...)

$predict(x, ...)

Makes predictions on new data (after model was trained)

Arguments

x

A matrix like object, should inherit from Matrix or matrix. Allowed classes should be defined in child classes.

y

target - usually vector, but also can be a matrix like object. Allowed classes should be defined in child classes.

...

additional parameters with default values

Base abstract class for all transformations

Description

Base class for all online transformations.

Usage

mlapiTransformation

Format

R6Class object.

Methods

$fit_transform(x, y = NULL, ...)

$transform(x, ...)

Performs transformation of the new data (after model was trained)

Arguments

x

A matrix like object, should inherit from Matrix or matrix. Allowed classes should be defined in child classes.

y

NULL. Optional taget variable. Usually this should be NULL. There few cases when it could be used.

...

additional parameters with default values

Base abstract class for all transformations which can be trained incremendally (online)

Description

Base class for all online transformations. This class inherits from mlapiTransformation and additionally requires to implement $partial_fit(x, y, ...) method. Idea is that user can pass x, y in chunks and model will be updated/refined incrementally.

Usage

mlapiTransformationOnline

Format

R6Class object.

Methods

$fit_transform(x, y = NULL, ...)

$transform(x, ...)

Performs transformation of the new data (after model was trained)

Arguments

x

A matrix like object, should inherit from Matrix or matrix. Allowed classes should be defined in child classes.

y

NULL. Optional taget variable. Usually this should be NULL. There few cases when it could be used.

...

additional parameters with default values

Makes predictions on new data using pre-trained model

Description

Makes predictions on new data using pre-trained model (inherits from mlapiEstimation)

Usage

## S3 method for class 'matrix'
predict(object, model, ...)

## S3 method for class 'Matrix'
predict(object, model, ...)

Arguments

Transforms new data using pre-trained model

Description

Generic function to transform data with pre-trained model (inherits from mlapiTransformation)

Usage

## S3 method for class 'Matrix'
transform(`_data`, model, ...)

## S3 method for class 'matrix'
transform(`_data`, model, ...)

Arguments