International Business Machines Corporation

Soumya Ghosh

Ying Li

Jianying Hu

Yu Cheng

Yasaman Khazaeni

Zhaonan Sun

Mayank Agarwal

Mikhail Yurochkin

Trong Nghia Hoang

Kristjan Herbert Greenewald

In federated learning problems, data is scattered across different servers and exchanging or pooling it is often impractical or prohibited. A Bayesian nonparametric framework is presented for federated learning with neural networks. Each data server is assumed to provide local neural network weights, which are modeled through our framework. An inference approach is presented that allows us to synthesize a more expressive global network without additional supervision, data pooling and with as few as a single communication round. The efficacy of the present invention on federated learning problems simulated from two popular image classification datasets is shown.

Bayesian nonparametric learning of neural networks

A system (or method) for generation and employment of disease progression model(s) that facilitates identifying and indexing discriminative features for disease progression in observational data. The disease progression prediction system comprises a processor that executes computer executable components stored in memory. A receiving component receives and learns observational patient data. A model generation component builds a preliminary disease progression model. An identification component identifies discriminative clinical features for different disease stages. A ranking component ranks discriminative powers of clinical features for respective pairs of disease stages; wherein the model generation component employs the ranked features to generate a final disease progression model.

Identifying and indexing discriminative features for disease progression in observational data

Growing community of inventors

Boston, MA, United States of America

Soumya Ghosh