Package 'distcomp'

Description

The function availableComputations returns a list of available computations with various components. The names of this list (with no spaces) are unique canonical tags that are used throughout the package to unambiguously refer to the type of computation; web applications particularly rely on this list to instantiate objects. As more computations are implemented, this list is augmented.

Usage

availableComputations()

Value

a list with the components corresponding to a computation

See Also

getComputationInfo()

Examples

availableComputations()

Description

The function availableDataSources returns the currently implemented data sources such as CSV files, Redcap etc.

Usage

availableDataSources()

Value

a list of named arguments, each of which is another list, with required fields named desc, a textual description and requiredPackages

Examples

availableDataSources()

Description

CoxMaster objects instantiate and run a distributed Cox model computation fit

Methods

Public methods

• CoxMaster$new()

• CoxMaster$kosher()

• CoxMaster$logLik()

• CoxMaster$addSite()

• CoxMaster$run()

• CoxMaster$summary()

• CoxMaster$clone()

Method new()

CoxMaster objects instantiate and run a distributed Cox model computation fit

Usage

CoxMaster$new(defn, debug = FALSE)

Arguments

Returns

R6 CoxMaster object

Method kosher()

Check if inputs and state of object are sane. For future use

Usage

CoxMaster$kosher()

Returns

TRUE or FALSE

Method logLik()

Return the partial log likelihood on all data for given beta parameter.

Usage

CoxMaster$logLik(beta)

Arguments

Returns

a named list with three components: value contains the value of the log likelihood, gradient contains the score vector, and hessian contains the estimated hessian matrix

Method addSite()

Add a url or worker object for a site for participating in the distributed computation. The worker object can be used to avoid complications in debugging remote calls during prototyping.

Usage

CoxMaster$addSite(name, url = NULL, worker = NULL)

Arguments

Method run()

Run the distributed Cox model fit and return the estimates

Usage

CoxMaster$run(control = coxph.control())

Arguments

Returns

a named list of beta, var, gradient, iter, and returnCode #' @description ' Return the summary of fit as a data frame

Method summary()

Usage

CoxMaster$summary()

Returns

a summary data frame columns for coef, exp(coef), ' standard error, z-score, and p-value for each parameter in the model following the same format as the survival package

Method clone()

The objects of this class are cloneable with this method.

Usage

CoxMaster$clone(deep = FALSE)

Arguments

See Also

CoxWorker which generates objects matched to such a master object

Description

CoxWorker objects are worker objects at each data site of a distributed Cox model computation

Methods

Public methods

• CoxWorker$new()

• CoxWorker$getP()

• CoxWorker$getStateful()

• CoxWorker$logLik()

• CoxWorker$var()

• CoxWorker$kosher()

• CoxWorker$clone()

Method new()

Create a new CoxWorker object.

Usage

CoxWorker$new(defn, data, stateful = TRUE)

Arguments

Returns

a new CoxWorker object

Method getP()

Return the dimension of the parameter vector.

Usage

CoxWorker$getP(...)

Arguments

Returns

the dimension of the parameter vector

Method getStateful()

Return the stateful status of the object.

Usage

CoxWorker$getStateful()

Returns

the stateful flag, TRUE or FALSE

Method logLik()

Return the partial log likelihood on local data for given beta parameter.

Usage

CoxWorker$logLik(beta, ...)

Arguments

Returns

a named list with three components: value contains the value of the log likelihood, gradient contains the score vector, and hessian contains the estimated hessian matrix

Method var()

Return the variance of estimate for given beta parameter on local data.

Usage

CoxWorker$var(beta, ...)

Arguments

Returns

variance vector

Method kosher()

Check if inputs and state of object are sane. For future use

Usage

CoxWorker$kosher()

Returns

TRUE or FALSE

Method clone()

The objects of this class are cloneable with this method.

Usage

CoxWorker$clone(deep = FALSE)

Arguments

See Also

CoxMaster which goes hand-in-hand with this object

Description

The function createHEWorkerInstance uses a definition identified by defnId to create the appropriate object instance for HE computations. The instantiated object is searched for in the instance path and loaded if already present, otherwise it is created and assigned the instanceId and saved under the dataFileName if the latter is specified. This instantiated object may change state between iterations when a computation executes

Usage

createHEWorkerInstance(
  defnId,
  instanceId,
  pubkey_bits = NULL,
  pubkey_n = NULL,
  den_bits = NULL,
  dataFileName = NULL
)

Arguments

Value

TRUE if everything goes well

See Also

availableComputations()

Description

This function uses an identifier (defnId) to locate a stored definition in the workspace to create the appropriate object instance. The instantiated object is assigned the instanceId and saved under the dataFileName if the latter is not NULL. This instantiated object may change state between iterations when a computation executes

Usage

createNCPInstance(
  name,
  ncpId,
  instanceId,
  pubkey_bits,
  pubkey_n,
  den_bits,
  dataFileName = NULL
)

Arguments

Value

TRUE if everything goes well

Description

The function createWorkerInstance uses a definition identified by defnId to create the appropriate object instance. The instantiated object is assigned the instanceId and saved under the dataFileName if the latter is specified. This instantiated object may change state between iterations when a computation executes

Usage

createWorkerInstance(
  defnId,
  instanceId,
  pubkey_bits = NULL,
  pubkey_n = NULL,
  den_bits = NULL,
  dataFileName = NULL
)

Arguments

Value

TRUE if everything goes well

See Also

availableComputations()

Description

This function just calls runDistcompApp() with the parameter "definition"

Usage

defineNewComputation()

Value

the results of running the web application

See Also

runDistcompApp()

Description

The function destroyInstanceObject deletes an object associated with the instanceId. This is typically done after a computation completes and results have been obtained.

Usage

destroyInstanceObject(instanceId)

Arguments

Value

TRUE if everything goes well

See Also

createWorkerInstance()

Description

distcomp is a collection of methods to fit models to data that may be distributed at various sites. The package arose as a way of addressing the issues regarding data aggregation; by allowing sites to have control over local data and transmitting only summaries, some privacy controls can be maintained. Even when participants have no objections in principle to data aggregation, it may still be useful to keep data local and expose just the computations. For further details, please see the reference cited below.

Details

The initial implementation consists of a stratified Cox model fit with distributed survival data and a Singular Value Decomposition of a distributed matrix. General Linear Models will soon be added. Although some sanity checks and balances are present, many more are needed to make this truly robust. We also hope that other methods will be added by users.

We make the following assumptions in the implementation: (a) the aggregate data is logically a stacking of data at each site, i.e., the full data is row-partitioned into sites where the rows are observations; (b) Each site has the package distcomp installed and a workspace setup for (writeable) use by the opencpu server (see distcompSetup(); and (c) each site is exposing distcomp via an opencpu server.

The main computation happens via a master process, a script of R code, that makes calls to distcomp functions at worker sites via opencpu. The use of opencpu allows developers to prototype their distributed implementations on a local machine using the opencpu package that runs such a server locally using localhost ports.

Note that distcomp computations are not intended for speed/efficiency; indeed, they are orders of magnitude slower. However, the models that are fit are not meant to be recomputed often. These and other details are discussed in the paper mentioned above.

The current implementation, particularly the Stratified Cox Model, makes direct use of code from survival::coxph(). That is, the underlying Cox model code is derived from that in the R survival survival package.

For an understanding of how this package is meant to be used, please see the documented examples and the reference.

References

Software for Distributed Computation on Medical Databases: A Demonstration Project. Journal of Statistical Software, 77(13), 1-22. doi:10.18637/jss.v077.i13

Appendix E of Modeling Survival Data: Extending the Cox Model by Terry M. Therneau and Patricia Grambsch. Springer Verlag, 2000.

See Also

The examples in system.file("doc", "examples.html", package="distcomp")

The source for the examples: system.file("doc_src", "examples.Rmd", package="distcomp").

Description

The function distcompSetup sets up a distributed computation and configures some global parameters such as definition file names, data file names, instance object file names, and ssl configuration parameters. The function creates some of necessary subdirectories if not already present and throws an error if the workspace areas are not writeable

Usage

distcompSetup(
  workspacePath = "",
  defnPath = paste(workspacePath, "defn", sep = .Platform$file.sep),
  instancePath = paste(workspacePath, "instances", sep = .Platform$file.sep),
  defnFileName = "defn.rds",
  dataFileName = "data.rds",
  instanceFileName = "instance.rds",
  resultsCacheFileName = "results_cache.rds",
  ssl_verifyhost = 1L,
  ssl_verifypeer = 1L
)

Arguments

Value

TRUE if all is well

See Also

getConfig()

Examples

## Not run: 
distcompSetup(workspacePath="./workspace")

## End(Not run)

Description

The function executeHEMethod is a homomorphic encryption wrapper around executeMethod. It ensures any returned result is encrypted using the homomorphic encryption function.

Usage

executeHEMethod(objectId, method, ...)

Arguments

Value

a list containing an integer and a fractional result converted to characters

Description

The function executeMethod is really the heart of distcomp. It executes an arbitrary method on an object that has been serialized to the distcomp workspace with any specified arguments. The result, which is dependent on the computation that is executed, is returned. If the object needs to save state between iterations on it, it is automatically serialized back for the ensuing iterations

Usage

executeMethod(objectId, method, ...)

Arguments

Value

a result that depends on the computation being executed

Description

A hash is generated based on the contents of the object

Usage

generateId(object, algo = "xxhash64")

Arguments

Value

the hash as a string

See Also

digest::digest()

Description

In distcomp, several web applications need to communicate between themselves. Since only one application is expected to be active at any time, they do so via a global store, essentially a hash table. This function retrieves the value of a name

Usage

getComputationInfo(name)

Arguments

Value

the value for the variable, NULL if not set

See Also

setComputationInfo()

Description

The function getConfig returns the values of the configuration parameters set up by distcompSetup

Usage

getConfig(...)

Arguments

Value

a list consisting of

See Also

distcompSetup()

Examples

## Not run: 
getConfig()

## End(Not run)

Description

HEMaster objects run a distributed computation based upon a definition file that encapsulates all information necessary to perform a computation. A master makes use of two non-cooperating parties which communicate with sites that perform the actual computations using local data.

Public fields

Methods

Public methods

• HEMaster$new()

• HEMaster$getNC_party()

• HEMaster$getPubkey()

• HEMaster$addNCP()

• HEMaster$run()

• HEMaster$clone()

Method new()

Create a HEMaster object to run homomorphic encrypted computation

Usage

HEMaster$new(defn)

Arguments

Returns

a HEMaster object

Method getNC_party()

Return a list of noncooperating parties (NCPs)

Usage

HEMaster$getNC_party()

Returns

a named list of length 2 of noncooperating party information

Method getPubkey()

Return the public key from the public private key pair

Usage

HEMaster$getPubkey()

Returns

an R6 Pubkey object

Method addNCP()

Add a noncooperating party to this master either using a url or an object in session for prototyping

Usage

HEMaster$addNCP(ncp_defn, url = NULL, ncpWorker = NULL)

Arguments

Method run()

Run a distributed homomorphic encrypted computation and return the result

Usage

HEMaster$run(debug = FALSE)

Arguments

Returns

the result of the distributed homomorphic computation

Method clone()

The objects of this class are cloneable with this method.

Usage

HEMaster$clone(deep = FALSE)

Arguments

See Also

NCP()

Create a homomorphic computation query count master object to employ worker objects generated by HEQueryCountWorker()

Description

HEQueryCountMaster objects instantiate and run a distributed homomorphic query count computation; they're instantiated by non-cooperating parties (NCPs)

Super class

distcomp::QueryCountMaster -> HEQueryCountMaster

Public fields

Methods

Public methods

• HEQueryCountMaster$new()

• HEQueryCountMaster$setParams()

• HEQueryCountMaster$kosher()

• HEQueryCountMaster$queryCount()

• HEQueryCountMaster$cleanup()

• HEQueryCountMaster$run()

• HEQueryCountMaster$clone()

Method new()

Create a new HEQueryCountMaster object.

Usage

HEQueryCountMaster$new(defn, partyNumber, debug = FALSE)

Arguments

Returns

a new HEQueryCountMaster object

Method setParams()

Set some parameters of the HEQueryCountMaster object for homomorphic computations

Usage

HEQueryCountMaster$setParams(pubkey_bits, pubkey_n, den_bits)

Arguments

Method kosher()

Check if inputs and state of object are sane. For future use

Usage

HEQueryCountMaster$kosher()

Returns

TRUE or FALSE

Method queryCount()

Run the distributed query count, associate it with a token, and return the result

Usage

HEQueryCountMaster$queryCount(token)

Arguments

Returns

the partial result as a list of encrypted items with components int and frac

Method cleanup()

Cleanup the instance objects

Usage

HEQueryCountMaster$cleanup()

Method run()

Run the homomorphic encrypted distributed query count computation

Usage

HEQueryCountMaster$run(token)

Arguments

Returns

the partial result as a list of encrypted items with components int and frac

Method clone()

The objects of this class are cloneable with this method.

Usage

HEQueryCountMaster$clone(deep = FALSE)

Arguments

See Also

HEQueryCountWorker() which goes hand-in-hand with this object

Create a homomorphic computation query count worker object for use with master objects generated by HEQueryCountMaster()

Description

HEQueryCountWorker objects are worker objects at each site of a distributed query count model computation using homomorphic encryption

Super class

distcomp::QueryCountWorker -> HEQueryCountWorker

Public fields

Methods

Public methods

• HEQueryCountWorker$new()

• HEQueryCountWorker$setParams()

• HEQueryCountWorker$queryCount()

• HEQueryCountWorker$clone()

Method new()

Create a new HEQueryMaster object.

Usage

HEQueryCountWorker$new(
  defn,
  data,
  pubkey_bits = NULL,
  pubkey_n = NULL,
  den_bits = NULL
)

Arguments

Returns

a new HEQueryMaster object

Method setParams()

Set some parameters for homomorphic computations

Usage

HEQueryCountWorker$setParams(pubkey_bits, pubkey_n, den_bits)

Arguments

Method queryCount()

Run the query count on local data and return the appropriate encrypted result to the party

Usage

HEQueryCountWorker$queryCount(partyNumber, token)

Arguments

Returns

the count as a list of encrypted items with components int and frac

Method clone()

The objects of this class are cloneable with this method.

Usage

HEQueryCountWorker$clone(deep = FALSE)

Arguments

See Also

HEQueryCountMaster() which goes hand-in-hand with this object

Description

The function makeDefinition returns a computational definition based on current inputs (from the global store) given a canonical computation type tag. This is a utility function for web applications to use as input is being gathered

Usage

makeDefinition(compType)

Arguments

Value

a data frame corresponding to the computation type

See Also

availableComputations()

Examples

## Not run: 
makeDefinition(names(availableComputations())[1])

## End(Not run)

Description

Instantiate a master process for HE operations

Usage

makeHEMaster(defn)

Arguments

Value

an master object for HE operations

Description

The function makeMaster returns a master object corresponding to the definition. The types of master objects that can be created depend upon the available computations

Usage

makeMaster(defn, partyNumber = NULL, debug = FALSE)

Arguments

Value

a master object of the appropriate class based on the definition

See Also

availableComputations()

Description

Instantiate an noncooperating party

Usage

makeNCP(
  ncp_defn,
  comp_defn,
  sites = list(),
  pubkey_bits = NULL,
  pubkey_n = NULL,
  den_bits = NULL
)

Arguments

Value

an NCP object

Description

The function makeWorker returns an object of the appropriate type based on a computation definition and sets the data for the object. The types of objects that can be created depend upon the available computations

Usage

makeWorker(defn, data, pubkey_bits = NULL, pubkey_n = NULL, den_bits = NULL)

Arguments

Value

a worker object of the appropriate class based on the definition

See Also

availableComputations()

Description

NCP objects are worker objects that separate a master process from communicating directly with the worker processes. Typically two such are needed for a distributed homomorphic computation. A master process can communicate with NCP objects and the NCP objects can communicate with worker processes. However, the two NCP objects, designated by numbers 1 and 2, are non-cooperating in the sense that they don't communicate with each other and are isolated from each other.

Public fields

Methods

Public methods

• NCP$new()

• NCP$getStateful()

• NCP$setParams()

• NCP$getSites()

• NCP$setSites()

• NCP$addSite()

• NCP$cleanupInstance()

• NCP$run()

• NCP$clone()

Method new()

Create a new NCP object.

Usage

NCP$new(
  ncp_defn,
  comp_defn,
  sites = list(),
  pubkey_bits = NULL,
  pubkey_n = NULL,
  den_bits = NULL
)

Arguments

Returns

a new NCP object

Method getStateful()

Retrieve the value of the stateful field

Usage

NCP$getStateful()

Method setParams()

Set some parameters of the NCP object for homomorphic computations

Usage

NCP$setParams(pubkey_bits, pubkey_n, den_bits)

Arguments

Method getSites()

Retrieve the value of the private sites field

Usage

NCP$getSites()

Method setSites()

Set the value of the private sites field

Usage

NCP$setSites(sites)

Arguments

Method addSite()

Add a url or worker object for a site for participating in the distributed computation. The worker object can be used to avoid complications in debugging remote calls during prototyping.

Usage

NCP$addSite(name, url = NULL, worker = NULL)

Arguments

Method cleanupInstance()

Clean up by destroying instance objects created in workspace.

Usage

NCP$cleanupInstance(token)

Arguments

Method run()

Run the distributed homomorphic computation

Usage

NCP$run(token)

Arguments

Returns

the result of the computation

Method clone()

The objects of this class are cloneable with this method.

Usage

NCP$clone(deep = FALSE)

Arguments

Create a master object to control worker objects generated by QueryCountWorker()

Description

QueryCountMaster objects instantiate and run a distributed query count computation

Methods

Public methods

• QueryCountMaster$new()

• QueryCountMaster$kosher()

• QueryCountMaster$queryCount()

• QueryCountMaster$getSites()

• QueryCountMaster$addSite()

• QueryCountMaster$run()

• QueryCountMaster$clone()

Method new()

Create a new QueryCountMaster object.

Usage

QueryCountMaster$new(defn, debug = FALSE)

Arguments

Returns

a new QueryCountMaster object

Method kosher()

Check if inputs and state of object are sane. For future use

Usage

QueryCountMaster$kosher()

Returns

TRUE or FALSE

Method queryCount()

Run the distributed query count and return the result

Usage

QueryCountMaster$queryCount()

Returns

the count

Method getSites()

Retrieve the value of the private sites field

Usage

QueryCountMaster$getSites()

Method addSite()

Add a url or worker object for a site for participating in the distributed computation. The worker object can be used to avoid complications in debugging remote calls during prototyping.

Usage

QueryCountMaster$addSite(name, url = NULL, worker = NULL)

Arguments

Method run()

Run the distributed query count

Usage

QueryCountMaster$run()

Returns

the count

Method clone()

The objects of this class are cloneable with this method.

Usage

QueryCountMaster$clone(deep = FALSE)

Arguments

See Also

QueryCountWorker() which goes hand-in-hand with this object

R6 worker object for use as a worker with master objects generated by QueryCountMaster()

Description

QueryCountWorker objects are worker objects at each site of a distributed QueryCount model computation

Methods

Public methods

• QueryCountWorker$new()

• QueryCountWorker$getStateful()

• QueryCountWorker$kosher()

• QueryCountWorker$queryCount()

• QueryCountWorker$clone()

Method new()

Create a new QueryCountWorker object.

Usage

QueryCountWorker$new(defn, data, stateful = FALSE)

Arguments

Returns

a new QueryCountWorker object

Method getStateful()

Retrieve the value of the stateful field

Usage

QueryCountWorker$getStateful()

Method kosher()

Check if inputs and state of object are sane. For future use

Usage

QueryCountWorker$kosher()

Returns

TRUE or FALSE

Method queryCount()

Return the query count on the local data

Usage

QueryCountWorker$queryCount()

Method clone()

The objects of this class are cloneable with this method.

Usage

QueryCountWorker$clone(deep = FALSE)

Arguments

See Also

QueryCountMaster() which goes hand-in-hand with this object

Description

In distcomp, several web applications need to communicate between themselves. Since only one application is expected to be active at any time, they do so via a global store, essentially a hash table. This function clears the store, except for the working directory.

Usage

resetComputationInfo()

Value

an empty list

See Also

setComputationInfo(), getComputationInfo()

Description

Web applications can define computation, setup worker sites or masters. This function invokes the appropriate web application depending on the task

Usage

runDistcompApp(appType = c("definition", "setupWorker", "setupMaster"))

Arguments

Value

the results of running the web application

See Also

defineNewComputation(), setupWorker(), setupMaster()

Description

The function saveNewComputation uses the computation definition to save a new computation instance. This is typically done for every site that wants to participate in a computation with its own local data. The function examines the computation definition and uses the identifier therein to uniquely refer to the computation instance at the site. This function is invoked (maybe remotely) on the opencpu server by uploadNewComputation() when a worker site is being set up

Usage

saveNewComputation(defn, data, dataFileName = NULL)

Arguments

Value

TRUE if everything goes well

See Also

uploadNewComputation()

Description

The function saveNewNCP uses the list of sites definition to save a new NCP instance. This is typically done for every pair of NCPs used in a computation. The function examines the computation definition and uses the identifier therein to uniquely refer to the computation instance at the site. This function is invoked (maybe remotely) on the opencpu server by uploadNewComputation() when a worker site is being set up

Usage

saveNewNCP(defn, comp_defn, data, dataFileName = NULL)

Arguments

Value

TRUE if everything goes well

See Also

uploadNewNCP()

Description

In distcomp, several web applications need to communicate between themselves. Since only one application is expected to be active at any time, they do so via a global store, essentially a hash table. This function sets a name to a value

Usage

setComputationInfo(name, value)

Arguments

Value

invisibly returns the all the name value pairs

See Also

getComputationInfo()

Description

This function just calls runDistcompApp() with the parameter "setupMaster"

Usage

setupMaster()

Value

the results of running the web application

See Also

runDistcompApp()

Description

This function just calls runDistcompApp() with the parameter "setupWorker"

Usage

setupWorker()

Value

the results of running the web application

See Also

runDistcompApp()

R6 class for SVD master object to control worker objects generated by SVDWorker()

Description

SVDMaster objects instantiate and run a distributed SVD computation

Methods

Public methods

• SVDMaster$new()

• SVDMaster$kosher()

• SVDMaster$updateV()

• SVDMaster$updateU()

• SVDMaster$fixFit()

• SVDMaster$reset()

• SVDMaster$addSite()

• SVDMaster$run()

• SVDMaster$summary()

• SVDMaster$clone()

Method new()

SVDMaster objects instantiate and run a distributed SVD computation

Usage

SVDMaster$new(defn, debug = FALSE)

Arguments

Returns

R6 SVDMaster object

Method kosher()

Check if inputs and state of object are sane. For future use

Usage

SVDMaster$kosher()

Returns

TRUE or FALSE

Method updateV()

Return an updated value for the V vector, normalized by arg

Usage

SVDMaster$updateV(arg)

Arguments

Returns

updated V

Method updateU()

Update U and return the updated norm of U

Usage

SVDMaster$updateU(arg)

Arguments

Returns

updated norm of U

Method fixFit()

Construct the residual matrix using given the V vector and d so far

Usage

SVDMaster$fixFit(v, d)

Arguments

Returns

result

Method reset()

Reset the computation state by initializing work matrix and set up starting values for iterating

Usage

SVDMaster$reset()

Method addSite()

Add a url or worker object for a site for participating in the distributed computation. The worker object can be used to avoid complications in debugging remote calls during prototyping.

Usage

SVDMaster$addSite(name, url = NULL, worker = NULL)

Arguments

Method run()

Run the distributed Cox model fit and return the estimates

Usage

SVDMaster$run(thr = 1e-08, max.iter = 100)

Arguments

Returns

a named list of V, d

Method summary()

Return the summary result

Usage

SVDMaster$summary()

Returns

a named list of V, d

Method clone()

The objects of this class are cloneable with this method.

Usage

SVDMaster$clone(deep = FALSE)

Arguments

See Also

SVDWorker() which goes hand-in-hand with this object

R6 class for a SVD worker object to use with master objects generated by SVDMaster()

Description

SVDWorker objects are worker objects at each site of a distributed SVD model computation

Methods

Public methods

• SVDWorker$new()

• SVDWorker$reset()

• SVDWorker$dimX()

• SVDWorker$updateV()

• SVDWorker$updateU()

• SVDWorker$normU()

• SVDWorker$fixU()

• SVDWorker$getN()

• SVDWorker$getP()

• SVDWorker$getStateful()

• SVDWorker$kosher()

• SVDWorker$clone()

Method new()

Create a new SVDWorker object.

Usage

SVDWorker$new(defn, data, stateful = TRUE)

Arguments

Returns

a new SVDWorker object

Method reset()

Reset the computation state by initializing work matrix and set up starting values for iterating

Usage

SVDWorker$reset()

Method dimX()

Return the dimensions of the matrix

Usage

SVDWorker$dimX(...)

Arguments

Returns

the dimension of the matrix

Method updateV()

Return an updated value for the V vector, normalized by arg

Usage

SVDWorker$updateV(arg, ...)

Arguments

Returns

updated V

Method updateU()

Update U and return the updated norm of U

Usage

SVDWorker$updateU(arg, ...)

Arguments

Returns

updated norm of U

Method normU()

Normalize U vector

Usage

SVDWorker$normU(arg, ...)

Arguments

Returns

TRUE invisibly

Method fixU()

Construct residual matrix using arg

Usage

SVDWorker$fixU(arg, ...)

Arguments

Method getN()

Getthe number of rows of x matrix

Usage

SVDWorker$getN()

Returns

the number of rows of x matrix

Method getP()

Getthe number of columnsof x matrix

Usage

SVDWorker$getP()

Returns

the number of columns of x matrix

Method getStateful()

Return the stateful status of the object.

Usage

SVDWorker$getStateful()

Returns

the stateful flag, TRUE or FALSE

Method kosher()

Check if inputs and state of object are sane. For future use

Usage

SVDWorker$kosher()

Returns

TRUE or FALSE

Method clone()

The objects of this class are cloneable with this method.

Usage

SVDWorker$clone(deep = FALSE)

Arguments

See Also

SVDMaster() which goes hand-in-hand with this object

Description

The function uploadNewComputation is really a remote version of saveNewComputation(), invoking that function on an opencpu server. This is typically done for every site that wants to participate in a computation with its own local data. Note that a site is always a list of at least a unique name element (distinguishing the site from others) and a url element.

Usage

uploadNewComputation(site, defn, data)

Arguments

Value

TRUE if everything goes well

See Also

saveNewComputation()

Description

The function uploadNewNCP is really a remote version of saveNewNCP(), invoking that function on an opencpu server. This is typically done for the two NCPs participating in a computation with the list of sites. Note that sites are always a list of at least a unique name element (distinguishing the site from others) and a url element.

Usage

uploadNewNCP(defn, comp_defn, url = NULL, worker = NULL, sites)

Arguments

Value

TRUE if everything goes well

See Also

saveNewNCP()

Description

Once a computation is defined, worker sites are set up, the master process code is written by this function. The current implementation does not allow one to mix localhost URLs with non-localhost URLs

Usage

writeCode(defn, sites, outputFilenamePrefix)

Arguments

Value

the value TRUE if all goes well

See Also

setupMaster()