Apple Inc.

International Business Machines Corporation

Ritwik K Kumar

Eric Daniel Friedman

Lucas Olen Winstrom

Tanveer Fathima Syeda-Mahmood

David James Beymer

Fei Wang

Augustin J Farrugia

Mathieu Ciet

Gianpaolo Fasoli

Yong Zhang

Karen Wolfe Brannon

Ting Chen

Benoit Chevallier-Mames

Julien Lerouge

Bruno Kindarji

Amol V Pattekar

Colin Compas

Gianluca Barbieri

Jefferson Provost

Vignesh Kumar

Jeremy M Stober

Gregory J Kuhlmann

Moritz A W Hardt

Tal Tversky

Eric M Gottschalk

Nathaniel C Bush

Systems and methods are disclosed for determining whether a message received by a client may be spam, in a computing environment that preserves privacy. The message may be encrypted. A client invokes the methods when a message is received from a sender that is not known to the client. A client can decrypt the message, break the message into chunks, and apply a differentially private algorithm to the set of chunks. The client transmits the differentially private message sketches to an aggregation server. The aggregation server receives a large collection of such message sketches for a large plurality of clients. The aggregation server returns aggregated message chunk (e.g. frequency) information to the client to assist the client in determining whether the message may be spam. The client can process the message based on the determination without disclosing the message content to the server.

Differential privacy for message text content mining

A user device can verify a user's identity to a server while protecting user privacy by not sharing personal data with any other device. To ensure user privacy, the user device performs an enrollment process in which the user performs an action sequence. The user device collects action data from the action sequence and uses the action data locally to generate a set of public/private key pairs (or other representation) from which information about the action sequence cannot be extracted. The public keys, but not the underlying action data, are sent to a server to store. To verify user identity, a user device can repeat the collection of action data and the generation of the key pairs. If the device can prove to the server its possession of the private keys to a sufficient degree, the user's identity can be verified.

Establishing and verifying identity using action sequences while protecting user privacy

A user device can verify a user's identity to a server while protecting user privacy by not sharing any personal data with any other device. To ensure user privacy and to allow multiple independent enrollments, the user device performs an enrollment process in which the user device locally collects and uses biometric data together with a random salt to generate a set of public/private key pairs from which biometric information cannot be extracted. The public keys and the salt, but not the biometric data, are sent to a server to store. To verify user identity, a user device can repeat the collection of biometric data from the user and the generation of public/private key pairs using the salt obtained from the server. If the device can prove to the server its possession of at least a minimum number of correct private keys, the user's identity can be verified.

Establishing and verifying identity using biometrics while protecting user privacy

Systems and methods are described for rate-limiting a message-sending client interacting with a message service based on dynamically calculated risk assessments of the probability that the client is, or is not, a sender of a spam messages. The message service sends a proof of work problem to a sending client device with a difficulty level that is related to a risk assessment that the client is a sender of spam messages. The message system limits the rate at which a known or suspected spammer can send messages by giving the known or suspected spammer client harder proof of work problems to solve, while minimizing the burden on normal users of the message system by given them easier proof of work problems to solve that can typically be solved by the client within the time that it takes to type a message.

Method for rate-limiting interactions based on dynamically calculated values by supplying problems of varying difficulty to be solved

Use of medical workflows where a first medical workflow is obtained from a plurality of medical acts performed in sequence that related to care of a patient. A set of condition-indication rules is applied to the first medical workflow to determine first condition information. The first condition information relates to a likelihood that a first medical condition exists in the patient.

Prediction of diseases based on analysis of medical exam and/or test workflow

Machine learning solutions compensate for data missing from input (training) data and thereby arrive at a predictive model that is based upon, and consistent with, the training data. The predictive model can be generated within a learning algorithm framework by transforming the training data to generate modality or similarity kernels. Similarity values can be generated for these missing similarity values.

Machine learning with incomplete data sets

Embodiments relate to segmenting blood vessels in angiogram images. An aspect includes a method that includes receiving and preprocessing at least one angiogram frame and preprocessing. In one embodiment, at least one angiogram frame is received and preprocessed. Bottom-up filtering of the preprocessed angiogram frame and top-down segmentation of the preprocessed angiogram frame are performed based on the results of the bottom-up filtering. The bottom-up filtering and the top-down segmentation are iteratively repeated until the difference between results of the top-down segmentation from consecutive iterations is equal to or below a threshold value. Based on determining that a difference between results of the top-down segmentation from consecutive iterations is below or equal to the threshold value, the results of the top-down segmentation are outputted.

Bidirectional blood vessel segmentation

According to one embodiment of the present invention, a method for echocardiogram view classification is provided. According to one embodiment of the present invention, a method comprises: obtaining a plurality of video images of a subject; aligning the plurality images; using the aligned images to generate a motion magnitude image; filtering the motion magnitude image using an edge map on image intensity; detecting features on the motion magnitude image, retaining only those features which lie in the neighborhood of intensity edges; encoding the remaining features by generating, x, y image coordinates, a motion magnitude histogram in a window around the feature point, and a histogram of intensity values near the feature point; and using the encoded features to classify the video images of the subject into a predetermined classification.

Echocardiogram view classification using edge filtered scale-invariant motion features

Growing community of inventors

San Jose, CA, United States of America

Ritwik K Kumar