Class: LBFGS

eclairjs/mllib/optimization. LBFGS

new LBFGS(gradient, updater)

Class used to solve an optimization problem using Limited-memory BFGS. Reference: http://en.wikipedia.org/wiki/Limited-memory_BFGS param: gradient Gradient function to be used. param: updater Updater to be used to update weights after every iteration.
Parameters:
Name Type Description
gradient module:eclairjs/mllib/optimization.Gradient
updater module:eclairjs/mllib/optimization.Updater
Source:

Methods

(static) runLBFGS(data, gradient, updater, numCorrections, convergenceTol, maxNumIterations, regParam, initialWeights, testData) → {module:eclairjs.Tuple2}

Run Limited-memory BFGS (L-BFGS) in parallel. Averaging the subgradients over different partitions is performed using one standard spark map-reduce in each iteration.
Parameters:
Name Type Description
data module:eclairjs.RDD - Input data for L-BFGS. RDD of the set of data examples, each of the form (label, [feature values]).
gradient module:eclairjs/mllib/optimization.Gradient - Gradient object (used to compute the gradient of the loss function of one single data example)
updater module:eclairjs/mllib/optimization.Updater - Updater function to actually perform a gradient step in a given direction.
numCorrections integer - The number of corrections used in the L-BFGS update.
convergenceTol float - The convergence tolerance of iterations for L-BFGS which is must be nonnegative. Lower values are less tolerant and therefore generally cause more iterations to be run.
maxNumIterations integer - Maximal number of iterations that L-BFGS can be run.
regParam float - Regularization parameter
initialWeights module:eclairjs/mllib/linalg.Vector (undocumented)
testData
Source:
Returns:
A tuple containing two elements. The first element is a column matrix containing weights for every feature, and the second element is an array containing the loss computed for every iteration.
Type
module:eclairjs.Tuple2

optimize(data, initialWeights) → {module:eclairjs/mllib/linalg.Vector}

Description copied from interface: Optimizer Solve the provided convex optimization problem.
Parameters:
Name Type Description
data module:eclairjs.RDD
initialWeights module:eclairjs/mllib/linalg.Vector
Source:
Returns:
Type
module:eclairjs/mllib/linalg.Vector

setConvergenceTol(tolerance) → {LBFGS}

Set the convergence tolerance of iterations for L-BFGS. Default 0.0001. Smaller value will lead to higher accuracy with the cost of more iterations. This value must be nonnegative. Lower convergence values are less tolerant and therefore generally cause more iterations to be run.
Parameters:
Name Type Description
tolerance float
Source:
Returns:
Type
LBFGS

setGradient(gradient) → {LBFGS}

Set the gradient function (of the loss function of one single data example) to be used for L-BFGS.
Parameters:
Name Type Description
gradient module:eclairjs/mllib/optimization.Gradient
Source:
Returns:
Type
LBFGS

setNumCorrections(corrections) → {LBFGS}

Set the number of corrections used in the LBFGS update. Default 10. Values of numCorrections less than 3 are not recommended; large values of numCorrections will result in excessive computing time. 3 < numCorrections < 10 is recommended. Restriction: numCorrections > 0
Parameters:
Name Type Description
corrections integer
Source:
Returns:
Type
LBFGS

setNumIterations(iters) → {LBFGS}

Set the maximal number of iterations for L-BFGS. Default 100.
Parameters:
Name Type Description
iters integer
Source:
Returns:
Type
LBFGS

setRegParam(regParam) → {LBFGS}

Set the regularization parameter. Default 0.0.
Parameters:
Name Type Description
regParam float
Source:
Returns:
Type
LBFGS

setUpdater(updater) → {LBFGS}

Set the updater function to actually perform a gradient step in a given direction. The updater is responsible to perform the update from the regularization term as well, and therefore determines what kind or regularization is used, if any.
Parameters:
Name Type Description
updater module:eclairjs/mllib/optimization.Updater
Source:
Returns:
Type
LBFGS