Microsoft Technology Licensing, LLC

Zhengwen Zhu

Nishant Velagapudi

Venkatasatya Premnath Ayyalasomayajula

Jason Craig Nelson

Lu Yang

David Nunez Tejerina

David Andrew Sterling

Siddhartha Mathur

Ozan Ozhan

Dionicio A Avila

Michael W Butler

Zhipeng Zhao

Rajkumar Ramasamy

Nicole Allen

Bradley Hughes

Shawn McGrath

John Tait

Aaron Whitney

Techniques performed by a data processing system for analyzing training data for a machine learning model and identifying outliers in the training data herein include obtaining training data for the model from a memory of the data processing system; analyzing the training data using a Siamese Neural Network to determine within-label similarities and cross-label similarities associated with a plurality of data elements within the training data, the within-label representing similarities between a respective data element and a first set of data elements similarly labeled in the training data, the cross-label similarities representing similarities between the respective data element and a second set of data elements dissimilarly labeled in the training data; identifying outlier data elements in the plurality of data elements based on the within-label and cross-label similarities; and processing the training data comprising the outlier data elements. Processing may include deleting the outlier data elements or generating a report.

Siamese neural networks for flagging training data in text-based machine learning

Techniques performed by a data processing system for analyzing the impact of training data changes on a machine learning model herein include training a first instance of a machine learning model with a first set of training data; modifying the first set of training data to produce a second set of training data; training a second instance of the model with the second set of training data; comparing the first instance of the model to the second instance of the model to determine features that differ between the first instance and the second instance of the model; identifying a subset of historical data associated with the features that differ between the first instance and the second instance of the model; and scoring the subset of the historical data to produce a report identifying differences in the output of the first instance and the second instance of the machine learning model.

Rapid adjustment evaluation for slow-scoring machine learning models

A method for dynamically generating a bookmark suggestion within a user interface of a computing device. The method includes accessing a target URL, receiving a request to bookmark the target URL, and determining if the target URL was accessed via a URL redirection function. The method also includes generating a prompt for display on the user interface. The prompt includes a user selectable option to save one of the target URL or a redirection URL associated with the target URL as a desired bookmark target address, based on the target URL being determined to have been accessed via the URL redirection function. The method further includes receiving an indication via the user interface of the desired bookmark target address, and saving the target URL or the redirection URL as the desired bookmark target address based on the received indication.

Dynamic bookmarking of web addresses accessed via short URL

Connectivity is tested using different locations and contexts selected from different possible failure zones. The failure zones may include: client failure zones; Internet failure zones; and online service failure zones. The results relating to different connectivity tests performed using the different failure zones are correlated and analyzed in an attempt to determine a root cause of the connectivity issue. For example, the root cause may be determined to be a configuration problem of the client, a problem with the client's networking equipment, an ISP problem, an Internet backbone problem; a problem of the online service, and the like. Different contexts may also be used when performing the tests. These results may be compared to the other connectivity test results. The results from the tests may be provided to the client experiencing the problem. Aggregated test results may also be used to detect service wide issues and trigger an alert.

Using different connectivity checks to determine causes of connectivity issues

The described implementations relate to processing of electronic data. One implementation is manifest as a system that can include logic and at least one processing device configured to execute the logic. The logic can be configured to receive a first task request to execute a first task that uses a resource when performed. The first task can have an associated first level of interactivity. The logic can also be configured to receive a second task request to execute a second task that also uses the resource when performed. The second task can have an associated second level of interactivity. The logic can also be configured to selectively throttle the first task and the second task based upon the first level of interactivity and the second level of interactivity.

Identifying and throttling tasks based on task interactivity

Growing community of inventors

Redmond, WA, United States of America

Zhengwen Zhu