International Business Machines Corporation

Sourav Mazumder

Shikhar Kwatra

Nadiya Kochura

Indervir Singh Banipal

Joseph N Kozhaya

Ryan R Anderson

Embodiments receive a data metrics input which corresponds with a microservice issue, determine a final outcome by multiplying each metric of the data metrics input by a weight, determine a link output based on the final outcome, determine a cost function based on the link output, evaluate that a cost output of the cost function has a minimized error, output a data metrics output based on the cost output of the cost function having the minimized error, train at least one machine learning (ML) model using the data metrics output, receive a new data metrics input which corresponds with another microservice issue, and provide a recommendation corresponding to the new data metrics input using the trained at least one ML model.

Alert management

An embodiment includes parsing an input dataset associated with a first node of a decision tree, where the input dataset includes a set of profile values for a set of projected usage parameters for a computing environment. The embodiment identifies a structure of the dataset using a recursive neural network that predicts a question sequence in a hierarchical tree format. The embodiment calculates a first deviation from the predicted question sequence and determines whether the deviation exceeds a threshold value. The embodiment generates a modified input dataset using a disambiguation rule and calculates a second deviation of the modified structure from the predicted question sequence and determines whether the deviation exceeds the threshold value. The embodiment assembles a customized hierarchical path using a generative model and assembles the customized hierarchical path by performing iterations of generating a series of candidate questions until a leaf node is reached.

Cognitive recommendation of computing environment attributes

A computer-implemented method for adapting a deep learning model to a local environment includes collecting training data; training a common deep learning model using the collected training data; customizing the deep learning model based on characteristics specific to one of a plurality of local devices utilizing transfer learning; and deploying the customized deep learning model to the one of the plurality of local devices. The customized deep learning model may be trained across a hierarchy comprising: a central common model trained on data from all devices in the plurality of devices; a location based model trained on a first set of data from devices in the plurality of devices within a location proximity; and a device based model trained on a data from one device in the plurality of devices.

Scalable and dynamic transfer learning mechanism

A computer-implemented method according to one embodiment includes building a knowledge graph of resource vendors. The method further includes identifying resources that are in relatively short supply in a host system, and determining, using a matching algorithm, a first of the resource vendors in the knowledge graph from which to temporarily lease the identified resources to the host system. The first resource vendor includes the identified resources and satisfies policy prerequisites of the host system. A lease is caused to be issued, from the first resource vendor to the host system, for the identified resources. A computer program product according to another embodiment includes a computer readable storage medium having program instructions embodied therewith. The program instructions are readable and/or executable by a computer to cause the computer to perform the foregoing method.

Knowledge graph for determining a resource vendor from which to lease resources

Generating a query response by receiving data for a non-user utterance, determining a question answering (QA) system response to the non-user utterance, receiving data for a user utterance responsive to the non-user utterance, determining a confidence score for the user utterance, determining a deviation between the user utterance and the QA system response, and providing the QA system response according to a combination of the deviation and the confidence score.

Augmenting user responses to queries

Disclosed are techniques for faceted search queries through interactive graphical objects. A text search query corresponding to a faceted search is received from a user, which is then parsed for keywords. Attributes associated with the keywords are then extracted and used to generate a graphical representation corresponding to the results associated with the text search query. Further user input to refine the query is provided through manipulation of the graphical representation by the user (such as dragging a leg of a table to change the length of legs on a table) corresponding to modification of one or more facets of the faceted search, narrowing the available search results corresponding to the faceted search.

Faceted search through interactive graphics

Approaches presented herein enable dynamically determining a validity of a return. More specifically, a system obtains a return request from a customer, a transaction history of the customer, and a set of return policy rules. A generative adversarial network (GAN) trained to detect non-genuine returns is applied to the return request. The GAN uses, among other this, the transaction history of the customer and the set of return policy rules as parameters of the GAN. Based on an output of the GAN, at least one return processing action is recommended and implemented.

Dynamic return optimization for loss prevention based on customer return patterns

Annotating programming code by receiving a first version of programming code, determining annotations for the code according to a first machine learning model, identifying a user, altering a generative model according a persona of the user, and generating altered annotations using the altered generative model.

Annotating program code

In an approach for initiating a chatbot session responsive to determining a product image and product description discrepancy, a processor, responsive to a user browsing a product listing for a product on an ecommerce website, processes a product description of the product listing using natural language processing (NLP) to generate a list of actor-subject negations from the product description. A processor processes a set of product images of the product listing to generate a list of entities from the set of product images. A processor correlates the list of actor-subject negations from the product description and the list of entities from the set of product images. A processor identifies at least one discrepancy between at least one actor-subject negation and at least one entity. A processor initiates, while the user is still browsing the product listing, a chat session with a chatbot of the ecommerce website.

Dynamic chatbot session based on product image and description discrepancy

A method, system and computer-usable medium are disclosed for improving likelihood of user to purchase a product or service. Interactions of a person, not necessarily the user, are monitored related to the product or service, along with actions by a business related to the person and the product or service. A trained Generative Adversarial Network model is applied to the monitored interactions to form recommend actions that the business should take to achieve ameliorative actions by the user. A reward feedback iterative adjustment of the GAN model is used to facilitate purchase of the product or service by the user.

System and method for generating preferred ameliorative actions using generative adversarial networks

A computer-implemented method, a computer program product, and a computer system for personalized machine learning model management and deployment on edge devices. An edge server monitors activities performed on respective ones of edge devices. The edge server associates machine learning models in a model set with respective ones of the activities. The edge server predicts a next set of activities that are to be performed on the respective ones of the edge devices. The edge server deploys, on the respective ones of edge devices, machine learning models that are associated with the next set of the activities. Applications on the respective ones of the devices, which execute the next set of the activities, leverage the machine learning models that are associated with the next set of the activities.

Personalized machine learning model management and deployment on edge devices

Aspects of the disclosure relate to Network Common Data Form (NetCDF) data management using a shared pool of configurable computing resources. A processing engine may be built to access a set of NetCDF data. The processing engine may be built by compiling a data frame software framework and compiling a Java library software framework. The set of NetCDF data may be received by the processing engine. A set of established data derived from the set of NetCDF data may be established using the processing engine. A set of operations on the set of established data may be processed. The processing may occur in a distributed fashion using the shared pool of configurable computing resources and the processing engine.

Growing community of inventors

San Ramon, CA, United States of America

Sourav Mazumder