International Business Machines Corporation

Patrick D Watson

Maria Chang

Jae-Wook Ahn

Aldis G Sipolins

Ravi Tejwani

Sharad C Sundararajan

Bishwaranjan Bhattacharjee

Parijat Dube

Ravindranath Kokku

John Ronald Kender

Hui Wu

Marco Cavallo

Jenna Reinen

Tengfei Ma

Lawrence Alfred Clevenger

Michael Robert Glass

Matthew Leon Hill

Noel C Codella

Siyu Huo

Brian Michael Belgodere

Chih-Chao Yang

Michael Rizzolo

Benjamin David Briggs

Christopher J Penny

Jayant R Kalagnanam

Nianjun Zhou

David S Wolpert

Prasenjit Dey

Dharmashankar Subramanian

Wesley M Gifford

Pavankumar Murali

Erwin Behnen

Timothy A Schell

Renuka Sindhgatta

Eric N Buseman

Shazia Afzal

Sharad Chandra Sundararajan

In an embodiment, a method for fine-tuning a pre-trained neural network for transfer learning, the method comprising obtaining a first target feature vector from a first layer of a pre-trained neural network responsive to a first target data element of a target dataset passing therethrough, obtaining a first source feature vector associated with the first layer of the pre-trained neural network, calculating a first divergence value for the first layer of the pre-trained neural network based at least in part on the first target feature vector and the first source feature vector, and setting a learning rate for the first layer of the pre-trained neural network based at least in part on the first divergence value.

Automated fine-tuning of a pre-trained neural network for transfer learning

A system includes a virtual reality system configured to enable a user to interact with a virtual environment, a plurality of biofeedback sensors configured to monitor a user, and a computer system including a virtual reality module configured to generate at least a view of the virtual environment, a biofeedback module configured to fuse output of the plurality of biofeedback sensors with a plurality of events within the virtual environment, a training module configured to generate a model of user behavior, wherein the training module executed by the computer system enables the computer system to make a prediction of a user response of the user based on a corpus of biofeedback data, and an alert module configured to generate at least one alert to the user via the virtual reality system based on the user response predicted by the computer system.

Building a model based on responses from sensors

Whether to train a new neural network model can be determined based on similarity estimates between a sample data set and a plurality of source data sets associated with a plurality of prior-trained neural network models. A cluster among the plurality of prior-trained neural network models can be determined. A set of training data based on the cluster can be determined. The new neural network model can be trained based on the set of training data.

Training transfer-focused models for deep learning

Techniques that facilitate layered stochastics anonymization of data are provided. In one example, a system includes a machine learning component and an evaluation component. The machine learning component performs a machine learning process for first data associated with one or more features to generate second data indicative of one or more example datasets within a degree of similarity to the first data. The first data and the second data comprise a corresponding data format. The evaluation component evaluates the second data for a particular feature from the one or more features and generates third data indicative of a confidence score for the second data.

Layered stochastic anonymization of data

A system and a method of managing a manufacturing process includes receiving production data relating to the manufacturing process and determining an operational mode associated with the manufacturing process using historical, multivariate senor data. The method may further determine a recommended action to affect production based on the determined operational mode. The operational mode may be based on at least one of: a level of operation in a continuous flow process relating to a joint set of process variables, a representation of a joint dynamic of the set of process variables over a predefined length, and a joint configuration of an uptime/downtime of a plurality of units comprising a process flow.

Generating mode change alerts with automatic detection from sensor data

Absorbent adhesive garment

Techniques for assessing the proficiency of artificial intelligence agents and users in a given knowledge domain are described. A plurality of proficiency agents can be initialized with a plurality of proficiency scores, by performing a plurality of assessments between pairs of proficiency agents selected from the plurality of proficiency agents. A first client device associated with a first user is matched with a first proficiency agent of the plurality of proficiency agents, based on a first proficiency score associated with the first user and a second proficiency score of the plurality of proficiency scores corresponding to the first proficiency agent. Assessments results of an assessment performed between the first client device and the first proficiency agent are received, and a rating system update function is used to update the first proficiency score and the second proficiency score, based on the assessment results.

Modeling mastery in distributed cognitive systems and storing the results in a ledger

Methods and systems of navigating within a virtual environment are described. In an example, a processor may generate a portal that includes a set of portal boundaries. The processor may display the portal within a first scene of the virtual environment being displayed on a device. The processor may display a second scene of the virtual environment within the portal boundaries. The processor may receive sensor data indicating a movement of a motion controller. The processor may reposition the portal and the second scene in the first scene based on the sensor data, wherein the first scene remains stationary on the device during the reposition of the portal and the second scene. The processor may translate a location of the portal within the first scene to move the portal towards a user of the device until the second scene replaces the first scene being displayed on the device.

Motion-controlled portals in virtual reality

A computer-implemented method for data labeling is provided. The computer-implemented method assigns pseudo-labels to unlabeled examples of data using a similarity metric on an embedding space to produce pseudo-labeled examples. A curriculum learning model is trained using the pseudo-labeled examples. The curriculum learning model trained with the pseudo-labeled examples is employed in in a fine-tuning task to enhance classification accuracy of the data.

Generating and augmenting transfer learning datasets with pseudo-labeled images

Systems, methods, and computer program products to perform an operation comprising receiving text input via a chat interface of a tutor application, identifying, by at least one classifier applied to the text input, a concept in the text input, mapping the concept in the text input to at least one of a visual action and a first visual object, generating, based on a first machine learning (ML) model, a first program code statement corresponding to the at least one of the visual action and the first visual object, and executing the first program code statement to modify a visualization interface of the tutor application based on the text input received via the chat interface.

Enhanced visual dialog system for intelligent tutors

A method and system of providing education content to a student is provided. A topic to be taught to a student is received. A user model of the student is identified. An electronic content model for the topic is determined. One or more presentation templates are identified based on the electronic content model. A concept map is determined based on the one or more presentation templates. A presentation template is selected from the one or more presentation templates. A segment of an educational content of the topic is provided to a user device of the student, based on the selected presentation template.

Adaptive presentation of educational content via templates

Method and apparatus for generating natural language observations using sequence-to-sequence models are provided. The method includes receiving a first electronic document and dynamically generating, without requiring user intervention, a first natural language observation corresponding to a first portion of the first electronic document by processing the first portion of the first electronic document using a first sequence-to-sequence model. A second natural language observation is received for the first portion of the first electronic document, and the generated first natural language observation and the authoritative natural language observation are provided. The method further includes receiving an evaluation of the first natural language observation and the second natural language observation, and refining the first sequence-to-sequence model based on the evaluation of the first natural language observation and second natural language observation.

Distributed system for evaluation and feedback of digital text-based content

A method for authenticating a user is presented. Responsive to a request for access to a computer resource, a computer system prompts the user making the request to access the computer resource to perform a new motion in an environment in which the user is monitored by a sensor system. Detected biometric data in the new motion performed by the user is identified by the computer system. A determination is made as to whether the user performing the new motion is an authenticated user based on comparing the detected biometric data with stored biometric data for a prior motion performed by the authenticated user. The computer system provides access to the computer resource when the user is identified as the authenticated user.

Motion based authentication

Systems, computer-implemented methods, and computer program products to generate virtual motion sensor data from computer animations are provided. 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 tracker component that can track virtual location data corresponding to a feature of a computer animated character in a virtual environment. The computer executable components can further comprise a virtual motion sensor component that, based on the virtual location data, can generate virtual motion sensor data.

Virtual generation of labeled motion sensor data

A method and system of interactive tutoring is provided. A question is sent to a user device. Upon determining that the response does not correspond to a known correct answer, a concept map is identified and a cost function between the correct answer and the response is determined. Upon determining that the cost function between the correct answer and the response is above a predetermined threshold, the question is identified to be the origin. Otherwise, the response is identified to be the origin. A present academic proficiency of the student is determined. A difficulty boundary based on the academic proficiency of the student is determined. A next question is provided between the origin and the correct answer, based on the determined difficulty boundary of the student.

Growing community of inventors

Ossining, NY, United States of America

Patrick D Watson