International Business Machines Corporation

Chen Lin

Maja Vukovic

Anup Kalia

John Joseph Rofrano

Jin Xiao

Jinho Hwang

Saurabh Sinha

Raghav Batta

Muhammed Fatih Bulut

Jie Ma

Ya Bin Dang

Nikolaos Anerousis

Shripad Nadgowda

Ali Kanso

Hongtan Sun

Komminist S Weldemariam

Li Jun Mei

Roxana Monge Nunez

Lijun Mei

Girmaw Abebe Tadesse

An approach to optimized migration of user assets to the cloud using artificial intelligence is presented. This approach may include user input and artificial intelligence trained with historical knowledge to generate rules. Migration models may be generated from the rules. A user may verify the migration models were successful. A task portfolio may be generated from the verified wave migration models. Runbook applications may be generated from the task portfolio and the migration may be executed using the runbooks.

Artificial intelligence optimized cloud migration

Systems and methods are provided that integrate a machine-learning model, and more specifically, utilizing a platform as a service (PaaS) cloud to predict probability of success for an operator in an environment. An embodiment comprises a system having: a processor that executes computer executable components stored in memory, trained machine-learning model that predicts probability of success for deployment of an operator in an environment with a namespace of a platform as a service (PaaS) cloud, and a deployment component that receives a first operator and a first namespace and employs the trained machine-learning model to predict success of deployment of the first operator in a first environment.

Machine-learning model to predict probability of success of an operator in a PaaS cloud environment

Systems, computer-implemented methods, and computer program products to facilitate generation of microservices from a monolithic application based on runtime traces are provided. According to an embodiment, a system can comprise a memory that stores computer executable components and a processor that executes the computer executable components stored in the memory. The computer executable components can comprise a model component that learns cluster assignments of classes in a monolithic application based on runtime traces of executed test cases. The computer executable components can further comprise a cluster component that employs the model component to generate clusters of the classes based on the cluster assignments to identify one or more microservices of the monolithic application.

Generation of microservices from a monolithic application based on runtime traces

Embodiments relate to a system, a program product, and a method to support a migration operation. Risk factors associated with a migration operation are assessed. The assessment includes assigning respective risk score values to the identified risk factors, assigning respective weight values to the identified risk factors, and calculating a composite risk score based on the assigned risk score values and the assigned weight values. At least one remediation action is recommended to reduce risk to the migration operation. The at least one recommended remediation action is implemented prior to execution of the migration operation. A migration plan incorporating the remediation action for the migration operation is generated. At least one machine learning (ML) model is employed in connection with (a) the subjecting of the identified risk factors to the assessment and/or (b) the recommending of at least one remediation action to reduce risk to the migration operation.

Migration risk assessment, recommendation, and implementation

Predicting infection risk by generating a first temporal graph of a first set of disease progression data, generating a second temporal graph of a second set of disease progression data, combining a first temporal graph node embedding and a second temporal graph node embedding, and generating a predicted infection risk according to the first temporal graph node embedding and the second temporal graph node embedding.

Predicting infection risk using heterogeneous temporal graphs

One embodiment of the invention provides a method for entity extraction, comprising determining a set of part-of-speech (POS) tags based on one or more documents, determining a concept in the one or more documents based on the set of POS tags, and extracting one or more phrases from the one or more documents based on the concept. The method further comprises generating a first set of rules corresponding to the concept based on the one or more phrases, generating a second set of rules specific to a domain based on the first set of rules, and learning, via an adapter grammar, a structure of one or more named entities in the one or more documents based on the second set of rules.

Unsupervised generation of rules for an adapter grammar

A system, program product, and method for use with an artificial intelligence (AI) platform to dynamically amend a knowledge base responsive to query evaluating and processing. A received or detected query is subject to natural language processing to identify, annotate, and map one or more query tokens against a knowledge base. The query tokens are evaluated against the knowledge base to identify one or more query tokens absent from the knowledge base and leverage a neural network to predict a probability relationship between the query tokens absent from the knowledge base and one or more tokens populated in the knowledge base. The natural language (NL) query is translated to a structured query language (SQL) and the SQL query is executed and evaluated, and the knowledge base is selectively and dynamically amended subject to the SQL evaluation.

Automated validity evaluation for dynamic amendment

Embodiments relate to a system, program product, and method for training a contrastive neural network (CNN) in an active learning environment. A neural network is pre-trained with labeled data of a historical (first) dataset. The CNN is trained for a new (second) dataset by applying the new dataset and contrasting the new dataset against the historical dataset to extract novel patterns. Weights of a knowledge operator from the pre-trained neural network are borrowed. Features novel to the new dataset are learned, including updating weights of the knowledge operator. The borrowed knowledge operator weights are combined with the updated knowledge operator weights. The CNN is leveraged to predict one or more labels for the new dataset as output data.

Contrastive neural network training in an active learning environment

A method, system, and computer program product for decomposing monolithic applications to form microservices are provided. The method identifies a set of classes within a monolithic application. A set of horizontal clusters are generated by performing horizontal clustering to the set of classes to decompose the classes based on a first functionality type. The method generates a set of vertical clusters by performing vertical clustering to the set of classes to decompose the classes based on a second functionality type. A subset of classes occurring in a common horizontal cluster and vertical cluster are identified as a functional unit. The method merges one or more functional units to form a microservice.

Forming microservices from monolithic applications

A computer system, computer program product, and computer-implemented method to identify one or more re-factoring operations directed at micro-service identification for source code. A genetic algorithm is leveraged to produce an offspring population of re-factoring operations from a parent set. The offspring population is subject to an assessment utilizing one or more objective measures. Responsive to the assessment, one or more identified re-factoring operations are selectively applied to the source code to produce one or more corresponding micro-service candidates.

Method and system to identify and prioritize re-factoring to improve micro-service identification

Systems and methods for prioritizing interview questions, including predicting answers to questions in a candidate question set using a neural network model, and generating a ranking for each of the questions in the candidate question set by determining answer confidence, completeness, and ambiguity for each of the answers, and incorporating user preferences using a personalized preference matrix. A top ranked question is automatically selected and presented to the user, and the personalized preference matrix is iteratively adjusted for subsequent ranking of questions based on an answer to the top ranked question by the user to reduce computational resources expended.

Predicting user preferences and requirements for cloud migration

Techniques for generating domain and client-specific application program interface recommendations are provided. In one example, a computer-implemented method comprises determining a semantic similarity between a first client application program interface and a second client application program interface based on a modified client application program interface being generated by removing text data associated with a first description of the first client application program interface. The computer-implemented method can further comprise matching the first description to a second description of the second client application program interface based on the semantic similarity, resulting in on or more matched client application program interfaces. Additionally, the computer-implemented method can further comprise associating the first client application program interface with a community based on the one or more matched client application program interfaces.

Facilitation of domain and client-specific application program interface recommendations

Techniques for generating domain and client-specific application program interface recommendations are provided. In one example, a computer-implemented method comprises modifying, by a device operatively coupled to a processor, a description of a client application program interface by removing text data associated with the description of the client application program interface, resulting in a modified description of the client application program interface. The computer-implemented method can further comprise analyzing, by the device, a performance associated with the client application program interface to generate an ontology based on a semantic similarity between the modified description of the client application program interface and one or more previous descriptions of one or more previous client application program interfaces.

Growing community of inventors

Elmsford, NY, United States of America

Chen Lin