Instance weighted learning machine learning model

1. A method of employing an instance weighted learning (IWL) machine learning model to train a classifier, the method comprising:

determining a quality value for each machine learning training instance in a temporal sequence of reinforcement learning machine learning training instances, each quality value being determined by determining a reward of a current machine learning training instance in the temporal sequence and determining a discounted portion of the reward that is added to each of the previous machine learning training instances in the temporal sequence, each of the machine learning training instances including a state-action pair;

associating the corresponding determined quality value with each of the machine learning training instances; and

training, using reinforcement learning, a classifier using each of the machine learning training instances, with each of the machine learning training instances weighted during the training based on its associated quality value using a weighting factor that is a function of its associated quality value, such that the training of the classifier is influenced more by a machine learning training instance with a higher quality value than by a machine learning training instance with a lower quality value.

2. The method as recited in claim 1 , wherein the classifier comprises a multilayer perceptron (MLP) neural network, another multilayer neural network, a decision tree, or a support vector machine.

3. The method as recited in claim 1 , wherein each of the quality values can be positive or negative, with a positive quality value tending to encourage learning to support instances similar to the associated machine learning training instance and a negative quality value tending to discourage learning to support instances similar to the corresponding machine learning training instance.

4. The method as recited in claim 1 , wherein each of the machine learning training instances is weighted during the training based on its associated quality value using a weighting factor that is a function of its associated quality value according to the following formula:

u ( q )=( a+b·q ), where:

q is the associated quality value;

u(q) is the weighting factor;

a is a first empirical parameter; and

b is a second empirical parameter.

5. The method as recited in claim 1 , wherein the discounted portion of the reward that is associated with each of the previous machine learning training instances is reduced the farther that each previous machine learning training instance is positioned in the temporal sequence from the current machine learning training instance.

6. The method as recited in claim 1 , wherein:

each of the machine learning training instances is a multiple output dependency (MOD) machine learning training instance, with each of the MOD machine learning training instances including multiple interdependent output components; and

training, using the reinforcement learning, the classifier using each of the MOD machine learning training instances includes employing a hierarchical based sequencing (HBS) machine learning model or a multiple output relaxation (MOR) machine learning model in the training.

7. The method as recited in claim 6 , wherein each of the MOD machine learning training instance is a lead response management (LRM) MOD machine learning training instance.

8. A non-transitory computer-readable medium storing a program configured to cause a processor to execute the method as recited in claim 1 .