ExpertMatrix: Expert Library Classes

In sparktseung/LRMoE: Logit-Weighted Reduced Mixture-of-Experts

Expert Library Classes

Description

This is the library of all the Expert Functions. In another word, it stored all the metadata of the Expert Functions. This object is a singleton, which means it should be only created when the package is loaded. It is also hidden in the global environment. Please do not clear the hidden objects And it you cleared the hidden objects. Please reload the package to fix that.

This is the matrix class of Expert Functions. Users are expected to perform all vector operations through ExpertMatrix to reduce index errors.

Public fields

continuous_experts

(NULL)
The names of continuous ExpertFunction

discrete_experts

(NULL)
The names of discrete ExpertFunction

distribution_names

(NULL)
The names of all expert functions Register a new Expert Function

Methods

Public methods

• ExpertLibrary$register()

• ExpertLibrary$is_continous()

• ExpertLibrary$is_discrete()

• ExpertLibrary$is_valid_expert()

• ExpertLibrary$clone()

---

Method register()

Usage

ExpertLibrary$register(expert_name, continuous)

Arguments

expert_name

(character)
The name of the expert function

continuous

(logical)
Is this new expert function continuous? Is given Expert Function continuous?

---

Method is_continous()

Usage

ExpertLibrary$is_continous(expert_name)

Arguments

expert_name

(character)
The name of the expert function Is given Expert Function discrete?

---

Method is_discrete()

Usage

ExpertLibrary$is_discrete(expert_name)

Arguments

expert_name

(character)
The name of the expert function Is given Expert Function got a correct name?

---

Method is_valid_expert()

Usage

ExpertLibrary$is_valid_expert(expert_name)

Arguments

expert_name

(character)
The name of the expert function

---

Method clone()

The objects of this class are cloneable with this method.

Usage

ExpertLibrary$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Public fields

nrow

(numeric)
The number of rows of ExpertMatrix

ncol

(numeric)
The number of columns of ExpertMatrix

expert_matrix

(matrix)
The matrix object to store all the ExpertFunction R6 objects

penalty_params

(matrix)
The matrix object to store all the penalty parameters for the Expert Functions

expert_params

(matrix)
The matrix object to store all the parameters for the Expert Functions Initialze a Expert Matrix Class

Methods

Public methods

• ExpertMatrix$new()

• ExpertMatrix$select()

• ExpertMatrix$exposurize()

• ExpertMatrix$get_mean()

• ExpertMatrix$get_variance()

• ExpertMatrix$set_penalty_params()

• ExpertMatrix$set_params()

• ExpertMatrix$get_penalty_value()

• ExpertMatrix$count_params()

• ExpertMatrix$clone()

---

Method new()

Usage

ExpertMatrix$new(expert_matrix, expert_params_matrix)

Arguments

expert_matrix

(matrix)
The name of Expert Functions, stored as a matrix of characters.

expert_params_matrix

(matrix)
The parameters of Expert Functions, stored as a matrix of lists. select the expert function object inside expert matrix

---

Method select()

Usage

ExpertMatrix$select(row_index = 0, col_index = 0)

Arguments

row_index

Row index

col_index

Column index Exposurize all the expert functions inside the expert matrix

---

Method exposurize()

Usage

ExpertMatrix$exposurize(exposure)

Arguments

exposure

The exposurized value

---

Method get_mean()

Return the mean matrix of the ExpertMatrix

Usage

ExpertMatrix$get_mean()

---

Method get_variance()

Return the variance matrix of the ExpertMatrix set the penalty parameters for all the expert functions

Usage

ExpertMatrix$get_variance()

---

Method set_penalty_params()

Usage

ExpertMatrix$set_penalty_params(expert_penalty_params_matrix)

Arguments

expert_penalty_params_matrix

(matrix) To be more specific, it is a matrix of list like this

penalty_params = matrix( list( c(2,1,2), c(1,1,1), c(3,1,2), c(1,3,4)), nrow = 2) set the parameters for all the expert functions

---

Method set_params()

Usage

ExpertMatrix$set_params(expert_params_matrix)

Arguments

expert_params_matrix

(matrix) To be more specific, it is a matrix of list of named list like this

params = matrix(list( list(meanlog = 1, sdlog = 2), list(meanlog = 1, sdlog = 2), list(meanlog = 3, sdlog = 1), list(meanlog = 5, sdlog = 1)), nrow = 2) Get the sum of penalty value of all the expert functions. Users are required to set the penalty parameters before call this function.

---

Method get_penalty_value()

Usage

ExpertMatrix$get_penalty_value(penalty)

Arguments

penalty

TRUE/FALSE

Returns

value The total penalty value of Expert Matrix Count the total number of parameters (for the Expert Functions) that exist in the Expert Matrix.

---

Method count_params()

Usage

ExpertMatrix$count_params()

Returns

result (integer) Total number of parameters

---

Method clone()

The objects of this class are cloneable with this method.

Usage

ExpertMatrix$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

params = matrix(list( list(meanlog = 1, sdlog = 2),  list(meanlog = 1, sdlog  = 2),
                      list(meanlog = 3, sdlog = 1),  list(meanlog = 5, sdlog = 1)), nrow = 2)
penalty_params = matrix( list( c(2,1,2), c(1,1,1), c(3,1,2), c(1,3,4)), nrow = 2)
expert_names = matrix( c("lognormal", "lognormal", "lognormal", "lognormal"), nrow = 2)
# ExpertMatrixObj = ExpertMatrix$new(expert_matrix = expert_names, expert_params_matrix = params)
# ExpertMatrixObj$set_penalty_params(penalty_params)
# ExpertMatrixObj$set_params(params)

sparktseung/LRMoE documentation built on March 21, 2022, 3:22 a.m.