Package ai.libs.jaicore.ml.ranking.dyad.learner.zeroshot.inputoptimization

Class PLNetInputOptimizer

java.lang.Object

ai.libs.jaicore.ml.ranking.dyad.learner.zeroshot.inputoptimization.PLNetInputOptimizer

public class PLNetInputOptimizer
extends java.lang.Object

Optimizes a given loss function (InputOptimizerLoss) with respect to the input of a PLNet using gradient descent. Assumes the PLNet was trained on normalized training data (i.e. scaled to intervals of 0 to 1 using DyadMinMaxScaler) and ensures that the optimized inputs will be within this range.

Constructor Summary

Constructors

Constructor	Description
PLNetInputOptimizer()

Method Summary

Modifier and Type	Method	Description
org.nd4j.linalg.api.ndarray.INDArray	optimizeInput(PLNetDyadRanker plNet, org.nd4j.linalg.api.ndarray.INDArray input, InputOptimizerLoss loss, double initialLearningRate, double finalLearningRate, int numSteps, org.nd4j.linalg.api.ndarray.INDArray inputMask)	Optimizes the given loss function with respect to a given PLNet's inputs using gradient descent.
org.nd4j.linalg.api.ndarray.INDArray	optimizeInput(PLNetDyadRanker plNet, org.nd4j.linalg.api.ndarray.INDArray input, InputOptimizerLoss loss, double initialLearningRate, double finalLearningRate, int numSteps, org.nd4j.linalg.primitives.Pair<java.lang.Integer,java.lang.Integer> indexRange)	Optimizes the given loss function with respect to a given PLNet's inputs using gradient descent.
org.nd4j.linalg.api.ndarray.INDArray	optimizeInput(PLNetDyadRanker plNet, org.nd4j.linalg.api.ndarray.INDArray input, InputOptimizerLoss loss, double learningRate, int numSteps, org.nd4j.linalg.api.ndarray.INDArray inputMask)	Optimizes the given loss function with respect to a given PLNet's inputs using gradient descent.
org.nd4j.linalg.api.ndarray.INDArray	optimizeInput(PLNetDyadRanker plNet, org.nd4j.linalg.api.ndarray.INDArray input, InputOptimizerLoss loss, double learningRate, int numSteps, org.nd4j.linalg.primitives.Pair<java.lang.Integer,java.lang.Integer> indexRange)	Optimizes the given loss function with respect to a given PLNet's inputs using gradient descent.
void	setListener(InputOptListener listener)	Set an InputOptListener to record the intermediate steps of the optimization procedure.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

PLNetInputOptimizer

public PLNetInputOptimizer()

Method Detail

optimizeInput

public org.nd4j.linalg.api.ndarray.INDArray optimizeInput(PLNetDyadRanker plNet,
                                                          org.nd4j.linalg.api.ndarray.INDArray input,
                                                          InputOptimizerLoss loss,
                                                          double learningRate,
                                                          int numSteps,
                                                          org.nd4j.linalg.primitives.Pair<java.lang.Integer,java.lang.Integer> indexRange)

Optimizes the given loss function with respect to a given PLNet's inputs using gradient descent. Ensures the outcome will be within the range of 0 and 1. Performs gradient descent for a given number of steps starting at a given input, using a static learning rate. The inputs that should be optimized can be specified using an index range in the form of a Pair} of integers.

Parameters:

plNet - PLNet whose inputs to optimize.

input - Initial inputs to start the gradient descent procedure from.

loss - The loss to be minimized.

learningRate - The initial learning rate.

numSteps - The number of steps to perform gradient descent for.

indexRange - Pair of indices (inclusive) specifying the parts of the input that should be optimized.

Returns:

The input optimized with respect to the given loss.

optimizeInput

public org.nd4j.linalg.api.ndarray.INDArray optimizeInput(PLNetDyadRanker plNet,
                                                          org.nd4j.linalg.api.ndarray.INDArray input,
                                                          InputOptimizerLoss loss,
                                                          double initialLearningRate,
                                                          double finalLearningRate,
                                                          int numSteps,
                                                          org.nd4j.linalg.primitives.Pair<java.lang.Integer,java.lang.Integer> indexRange)

Optimizes the given loss function with respect to a given PLNet's inputs using gradient descent. Ensures the outcome will be within the range of 0 and 1. Performs gradient descent for a given number of steps starting at a given input, using a linearly decaying learning rate. The inputs that should be optimized can be specified using an index range in the form of a Pair} of integers.

Parameters:

plNet - PLNet whose inputs to optimize.

input - Initial inputs to start the gradient descent procedure from.

loss - The loss to be minimized.

initialLearningRate - The initial learning rate.

finalLearningRate - The value the learning rate should decay to.

numSteps - The number of steps to perform gradient descent for.

indexRange - Pair of indices (inclusive) specifying the parts of the input that should be optimized.

Returns:

The input optimized with respect to the given loss.

optimizeInput

public org.nd4j.linalg.api.ndarray.INDArray optimizeInput(PLNetDyadRanker plNet,
                                                          org.nd4j.linalg.api.ndarray.INDArray input,
                                                          InputOptimizerLoss loss,
                                                          double learningRate,
                                                          int numSteps,
                                                          org.nd4j.linalg.api.ndarray.INDArray inputMask)

Optimizes the given loss function with respect to a given PLNet's inputs using gradient descent. Ensures the outcome will be within the range of 0 and 1. Performs gradient descent for a given number of steps starting at a given input, using a static learning rate. The inputs that should be optimized can be specified using a 0,1-vector

Parameters:

plNet - PLNet whose inputs to optimize.

input - Initial inputs to start the gradient descent procedure from.

loss - The loss to be minimized.

learningRate - The initial learning rate.

numSteps - The number of steps to perform gradient descent for.

inputMask - 0,1 vector specifying the inputs to optimize, i.e. should have a 1 at the index of any input that should be optimized and a 0 elsewhere.

Returns:

The input optimized with respect to the given loss.

optimizeInput

public org.nd4j.linalg.api.ndarray.INDArray optimizeInput(PLNetDyadRanker plNet,
                                                          org.nd4j.linalg.api.ndarray.INDArray input,
                                                          InputOptimizerLoss loss,
                                                          double initialLearningRate,
                                                          double finalLearningRate,
                                                          int numSteps,
                                                          org.nd4j.linalg.api.ndarray.INDArray inputMask)

Optimizes the given loss function with respect to a given PLNet's inputs using gradient descent. Ensures the outcome will be within the range of 0 and 1. Performs gradient descent for a given number of steps starting at a given input, using a linearly decaying learning rate. The inputs that should be optimized can be specified using a 0,1-vector

Parameters:

plNet - PLNet whose inputs to optimize.

input - Initial inputs to start the gradient descent procedure from.

loss - The loss to be minimized.

initialLearningRate - The initial learning rate.

finalLearningRate - The value the learning rate should decay to.

numSteps - The number of steps to perform gradient descent for.

inputMask - 0,1 vector specifying the inputs to optimize, i.e. should have a 1 at the index of any input that should be optimized and a 0 elsewhere.

Returns:

The input optimized with respect to the given loss.

setListener

public void setListener(InputOptListener listener)

Set an InputOptListener to record the intermediate steps of the optimization procedure.

Parameters:

listener -