International Business Machines Corporation

Massachusetts Institute of Technology

Subhro Das

Hung-Yang Chang

James V Codella

Chuang Gan

Dakuo Wang

Yang Zhang

Kenneth James Barker

Sijia Liu

Pei-Yun S Hsueh

Sara Rosenthal

Adam R Faulkner

Ching-Hua Chen

Prabhat Maddikunta Reddy

Chandramouli Maduri

Trong Nghia Hoang

Wyatt Clarke

Sebastian Steffen

Erik Brynjolfsson

Martin Fleming

Nathan Fulton

Nathan Hunt

Gema Almoguera

Obtain, as input, in electronic form, structured information including tasks for a plurality of occupations in a plurality of industries over a length of time; compute, from the structured information, a time series of normalized occupation task shares over the length of time; train a computerized machine learning model, on the time series, to predict future task shares for the plurality of occupations in the plurality of industries; and, with the trained computerized machine learning model, predict the future task shares.

Skills and tasks demand forecasting

A mechanism is provided in a data processing system to implement a personalized patient engagement engine. The personalized patient engagement engine develops a set of models for a plurality of behavioral phenotypes based on anonymized unstructured and structured patient-care management records for a plurality of patients over a period of time; matches a given patient to a behavioral phenotype; estimates a propensity of positive and/or negative behavioral responses of each of a plurality of targeted behaviors; dynamically updates personalized intervention effectiveness rankings in context for care manager and patient decision-making based on what has been shown to lead to positive responses for individuals with a similar behavioral profile; generates an intervention recommendation for the given patient based on the personalized intervention effectiveness rankings relative to the patient given an assigned goal and an individual intervention effect estimation; and provides the intervention recommendation to the care manager.

Personalized patient engagement in care management using explainable behavioral phenotypes

A method for training control software to reinforce safety constraints using visual inputs includes performing template matching for each object in an image of a reinforcement learning (RL) agent's action space using a visual template for each object wherein each object in the RL agent's action space is detected, mapping each detected object to a set of planar coordinates for each object in the RL agent's action space, determining a set of safe actions for the RL agent by applying a safety specification for the RL agent's action space to the set of variables for coordinates for each object in the RL agent's action space, outputting the set of safe actions to the RL agent for a current state of a RL procedure, and preventing the RL agent from executing an action that is unsafe, before the RL agent takes an action.

Formally safe symbolic reinforcement learning on visual inputs

Systems, computer-implemented methods, and computer program products to facilitate data protection distributed learning 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 local parameter component that employs an agent to compute local model parameters based on data of the agent. The computer executable components can further comprise a global parameter component that employs the agent to estimate a global model parameter based on the local model parameters and model parameters of one or more neighbor agents.

Data protection distributed learning

Method and apparatus that includes receiving a query describing an aspect in a video, the video including a plurality of frames, identifying multiple proposals that potentially correspond to the query where each of the proposals includes a subset of the plurality of frames, ranking the proposals using a graph convolution network that identifies relationships between the proposals, and selecting, based on the ranking, one of the proposals as a video segment that correlates to the query.

Graph convolutional networks for video grounding

Technical solutions are described that address correcting input time-series data provided for analysis and predictions. An example computer-implemented method includes receiving, by a processor, a time-series data input by a user. The computer-implemented method also includes computing, by the processor, a first plurality of predicted values based on the time-series data input by the user; computing, by the processor, a second plurality of predicted values by. The computer-implemented method also includes determining estimated time-series data based on the time-series data input by the user. The computer-implemented method also includes computing the second plurality of predicted values based on the estimated time-series data. The computer-implemented method also includes determining, by the processor, a defect in the time-series data input by the user based on a distribution of a plurality of differences between respective values from the first plurality of predicted values and the second plurality of predicted values.

Input data correction

Growing community of inventors

Cambridge, MA, United States of America

Subhro Das