Method and system for training of artificial intelligence and machine learning models

A method for interactively monitoring a training status in a machine learning model, the method being implemented by at least one processor, the method comprising:

• receiving, by the at least one processor, machine learning (ML) model training data, wherein the ML model training data includes an action-state pair, wherein the action-state pair relates to an action that results in a state transition;
• automatically generating, by the at least one processor via a training analyzer and based on the ML model training data, a first plurality of parameters that relate to measuring a training status of a first ML model;
• executing, by the at least one processor via an automatic training executive, a first training operation with respect to the first ML model in order to generate first respective values of each of the first plurality of parameters;
• receiving, by the at least one processor from a user, a first user input that relates to a first task that is performable by using the first ML model;
• performing, by the at least one processor, the first task based on the first user input;
• analyzing, by the at least one processor via the training analyzer, a result of the first training operation and a first impact upon the first ML model caused by the first user input;
• generating, by the at least one processor via a training adjustment artificial intelligence (AI) model and based on results of the first training operation, a plurality of training parameter sets for subsequent training operations,
• wherein the training analyzer and the training adjustment AI model adjust a hyperparameter and a reward function associated with the first training operation for implementing reinforced learning in the first training operation,
• wherein the reward function generates a reward for the action-state pair based on a proximity of the action-state pair to a predetermined outcome,
• wherein the hyperparameters use the reward as feedback to generate policies for selecting actions to produce the predetermined outcome, and
• wherein the implementing of the reinforced learning includes iteratively performing a series of training operations;
• executing, by the at least one processor via the automatic training executive, a second training operation with respect to a second ML model in order to generate second respective values for each of the first plurality of parameters;
• calculating, by the at least one processor, a first performance metric for the first ML model in relation to a performing of a predetermined task using the first ML model;
• calculating, by the at least one processor, a second performance metric for the second ML model in relation to a performing of the predetermined task using the second ML model;
• comparing, by the at least one processor, the first performance metric with the second performance metric;
• selecting, by the at least one processor, one from among the first ML model and the second ML model, based on a result of the comparing;
• and
• transmitting, by the at least one processor to the user, a first notification message that includes a result of the selection.

A method for monitoring a status of a machine learning model in a machine learning model life cycle, the method being implemented by at least one processor, the method comprising:

• receiving, by the at least one processor, machine learning (ML) model training data, wherein the ML model training data includes an action-state pair, wherein the action-state pair relates to an action that results in a state transition;
• automatically generating, by the at least one processor via a training analyzer and based on the ML model training data, a first plurality of parameters that relate to measuring a life cycle status of a first ML model;
• executing, by the at least one processor via an automatic training executive, a first life cycle development operation with respect to the first ML model in order to generate first respective values of each of the first plurality of parameters;
• analyzing, by the at least one processor via the training analyzer, a result of the first life cycle development operation based on the first respective values of each of the first plurality of parameters;
• generating, by the at least one processor via a training adjustment artificial intelligence (AI) model and based on results of the first life cycle development operation, a plurality of training parameter sets for subsequent training operations,
• wherein the training analyzer and the training adjustment AI model adjust a hyperparameter and a reward function associated with the first life cycle development operation for implementing reinforced learning in the first life cycle development operation,
• wherein the reward function generates a reward for the action-state pair based on a proximity of the action-state pair to a predetermined outcome,
• wherein the hyperparameters use the reward as feedback to generate policies for selecting actions to produce the predetermined outcome, and
• wherein the implementing of the reinforced learning includes iteratively performing a series of training operations;
• determining, by the at least one processor based on a result of the analyzing, whether the first life cycle development operation has generated a progression with respect to the ML model life cycle;
• executing, by the at least one processor via the automatic training executive, a second life cycle development operation with respect to a second ML model in order to generate second respective values for each of the first plurality of parameters;
• calculating, by the at least one processor, a first performance metric for the first ML model in relation to a performing of a predetermined task using the first ML model;
• calculating, by the at least one processor, a second performance metric for the second ML model in relation to a performing of the predetermined task using the second ML model;
• comparing, by the at least one processor, the first performance metric with the second performance metric;
• and
• selecting, by the at least one processor, one from among the first ML model and the second ML model, based on a result of the comparing.