SYSTEMS, METHODS, AND APPARATUSES FOR IMPLEMENTING MACHINE LEARNING MODEL TRAINING AND DEPLOYMENT WITH A ROLLBACK MECHANISM
First Claim
1. A method performed by a machine learning platform having at least a processor and a memory therein, wherein the method comprises:
- receiving training data as input at the machine learning platform, wherein the training data includes a multiple transactions, each of the transactions specifying a plurality of features upon which to make a prediction and a label representing a correct answer for the plurality of features according to each respective transaction;
specifying a model to be trained by the machine learning platform using the training data, wherein the model includes a plurality of algorithms and source code;
generating a new predictive engine variant by training the model to algorithmically arrive upon the label representing the correct answer as provided with the training data based on the plurality of features for each of the multiple transactions;
versioning the new predictive engine variant based at least on the time the new predictive engine variant was generated a version of the source code utilized within the model and the training data received as input;
deploying the new predictive engine variant into a production environment to replace a prior version of the predictive engine variant; and
rolling back the new predictive engine variant from the production environment to a specified version which is less than a version of the new predictive engine variant.
1 Assignment
0 Petitions
Accused Products
Abstract
In accordance with disclosed embodiments, there are provided systems, methods, and apparatuses for implementing machine learning model training and deployment with a rollback mechanism within a computing environment. For example, an exemplary machine learning platform includes means for receiving training data as input at the machine learning platform, in which the training data includes a multiple transactions, each of the transactions specifying a plurality of features upon which to make a prediction and a label representing a correct answer for the plurality of features according to each respective transaction; specifying a model to be trained by the machine learning platform using the training data, in which the model includes a plurality of algorithms and source code; generating a new predictive engine variant by training the model to algorithmically arrive upon the label representing the correct answer as provided with the training data based on the plurality of features for each of the multiple transactions; versioning the new predictive engine variant based at least on the time the new predictive engine variant was generated a version of the source code utilized within the model and the training data received as input; deploying the new predictive engine variant into a production environment to replace a prior version of the predictive engine variant; and rolling back the new predictive engine variant from the production environment to a specified version which is less than a version of the new predictive engine variant. Other related embodiments are disclosed.
-
Citations
25 Claims
-
1. A method performed by a machine learning platform having at least a processor and a memory therein, wherein the method comprises:
-
receiving training data as input at the machine learning platform, wherein the training data includes a multiple transactions, each of the transactions specifying a plurality of features upon which to make a prediction and a label representing a correct answer for the plurality of features according to each respective transaction; specifying a model to be trained by the machine learning platform using the training data, wherein the model includes a plurality of algorithms and source code; generating a new predictive engine variant by training the model to algorithmically arrive upon the label representing the correct answer as provided with the training data based on the plurality of features for each of the multiple transactions; versioning the new predictive engine variant based at least on the time the new predictive engine variant was generated a version of the source code utilized within the model and the training data received as input; deploying the new predictive engine variant into a production environment to replace a prior version of the predictive engine variant; and rolling back the new predictive engine variant from the production environment to a specified version which is less than a version of the new predictive engine variant. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16)
-
-
17. Non-transitory computer readable storage media having instructions stored thereupon that, when executed by a processor and memory of a machine learning platform, the instructions cause the machine learning platform to perform operations including:
-
receiving training data as input at the machine learning platform, wherein the training data includes a multiple transactions, each of the transactions specifying a plurality of features upon which to make a prediction and a label representing a correct answer for the plurality of features according to each respective transaction; specifying a model to be trained by the machine learning platform using the training data, wherein the model includes a plurality of algorithms and source code; generating a new predictive engine variant by training the model to algorithmically arrive upon the label representing the correct answer as provided with the training data based on the plurality of features for each of the multiple transactions; versioning the new predictive engine variant based at least on the time the new predictive engine variant was generated a version of the source code utilized within the model and the training data received as input; deploying the new predictive engine variant into a production environment to replace a prior version of the predictive engine variant; and rolling back the new predictive engine variant from the production environment to a specified version which is less than a version of the new predictive engine variant. - View Dependent Claims (18, 19, 20)
-
-
21. A machine learning platform comprising:
-
a memory to store instructions; a processor to execute instructions; a receive interface to receive training data as input at the machine learning platform, wherein the training data includes a multiple transactions, each of the transactions specifying a plurality of features upon which to make a prediction and a label representing a correct answer for the plurality of features according to each respective transaction; the receive interface to receive a specified model to be trained by the machine learning platform using the training data, wherein the model includes a plurality of algorithms and source code; a predictive engine generator to generate a new predictive engine variant by training the model to algorithmically arrive upon the label representing the correct answer as provided with the training data based on the plurality of features for each of the multiple transactions; a versioning system interfaced with the machine learning platform to version the new predictive engine variant based at least on the time the new predictive engine variant was generated a version of the source code utilized within the model and the training data received as input; a deployment platform to deploy the new predictive engine variant into a production environment to replace a prior version of the predictive engine variant; and the deployment platform to roll back the new predictive engine variant from the production environment to a specified version maintained by the versioning system which is less than a version of the new predictive engine variant. - View Dependent Claims (22, 23, 24, 25)
-
Specification