Dennis Wei

US Patent:

8493262, Jul 23, 2013

Filed:

Feb 11, 2011

Appl. No.:

13/026085

Inventors:

Petros T. Boufounos - Boston MA, US
Dennis Wei - Cambridge MA, US

Assignee:

Mitsubishi Electric Research Laboratories, Inc. - Cambridge MA

International Classification:

G01S 13/00

US Classification:

342 25R

Abstract:

A saturated input signal acquired by a synthetic aperture radar (SAR) system is processed by estimating a reconstruction that generated the input signal, reproducing an input signal from an estimated reconstruction to generate a reproduced signal, comparing the reproduced signal with the input signal; adjusting an estimated reconstruction based on the comparison; and iterating from the reproducing step until a termination condition is reached.

US Patent:

20230021338, Jan 26, 2023

Filed:

Jul 7, 2021

Appl. No.:

17/368925

Inventors:

- Armonk NY, US
Prasanna Sattigeri - Acton MA, US
Karthikeyan Shanmugam - Elmsford NY, US
Dennis Wei - Sunnyvale CA, US
Murat Kocaoglu - West Lafayette IN, US
Karthikeyan Natesan Ramamurthy - Pleasantville NY, US

International Classification:

G06N 3/08
G06N 3/04

Abstract:

A method for training a machine learning system using conditionally independent training data includes receiving an input dataset (p(x, y, z)). A generative adversarial network, that includes a generator and a first discriminator, uses the input dataset to generate a training data (p(x, y, z)) by generating the values (x, y, z). The first discriminator determines a first loss (L) based on (x, y, z) and (x, y, z). A divergence calculator modifies the training data based on a dependence measure (γ). The divergence calculator includes a second discriminator and a third discriminator. Modifying the training data includes receiving a reference value ({tilde over (y)}), and computing, by the second discriminator, a second loss (L) based on (x, y, z) and (x, {tilde over (y)}, z). The third discriminator computes a third loss (L) based on (y, z) and ({tilde over (y)}, z). Further, a fourth loss (L) is computed based on Land L. The training data is output from the generator if Land Lsatisfy a predetermined condition.