Apple Inc.

Gaurav Kapoor

Umesh Suresh Vaishampayan

Kit-Man Wan

Michael R Siracusa

Keith Alan Cox

Phillip Stanley-Marbell

Andrew H Vyrros

Abhradeep Guha Thakurta

Julien F Freudiger

Alexander Barraclough Brown

Stephen Olivier Lemay

Patrick L Coffman

Christopher Patrick Foss

Lawrence Yuan Yang

Richard R Dellinger

Xin Ru Wang

Jerome Rene Bellegarda

Cyrus Daniel Irani

Alexandre F Carlhian

Giulia Maria Pagallo

Daniel C Gross

Francesco Rossi

Thomas Deniau

Alexandre R Moha

Sofiane Toudji

Brent D Ramerth

Mathieu Jean Martel

Jason J Gauci

Colin C Morris

Jun Hatori

Karl Christian Kohlschuetter

Aria D Haghighi

Hafid J Arras

Kevin D Clark

Cyril De La Cropte De Chanterac

Vipul Ved Prakash

Jannes G A Dolfing

Bronwyn A Jones

Arnaud Legendre

Stephen C Peters

Steven Duplinsky

John Gregory Dorsey

David Michael Chan

Jonathan Jay Andrews

James S Ismail

Srikrishna Sridhar

Binu K Mathew

Zachary A Nation

Christopher M Hanson

Doug Davidson

Julien Freudinger

Jesper S Andersen

Elizabeth Ottens

Vrushali Mundhe

Vivek Rangarajan Sridhar

Matthew Elliot Shepherd

John Meron Ananny

Amit K Vyas

Kartik R Venkatraman

Anand Ramadurai

Varaprasad V Lingutla

Cecile Marie Foret

Lee Russell

David B Myszewski

William B March

Etienne Belanger

Brittany D Hughes

Michael F Culbert

Daniel Ben Pollack

Marc Jason Krochmal

Patrick Y Law

John Joseph Iarocci

Jay C Blahnik

Daniel S Keen

Joshua V Graessley

John M Hornkvist

Xiaojin Shi

Marco Zuliani

Vaughn Todd Arnold

Andrei Dorofeev

Abhishek Bhowmick

Alex J Crumlin

Ryan M Rogers

Joao Pedro Lacerda

Derek J DiCarlo

Jason Luo Pang

Stephen D Sterz

Daniel M Vinegrad

Joao Pedro Forjaz De Lacerda

Alexey Marinichev

Andrew M Matuschak

Jesper A Andersen

Justin N Wood

Vignesh Jagadeesh

Frank Faqiu Liang

Bronwyn Anne Jones

Amaury J Heresztyn

Nagarajan Kalyanasundaram

Sowmya Gopalan

Xin Wang

Systems and methods are disclosed for advising a user when an energy storage device in a computing system needs charging. State of charge data of the energy storage device can be measured and stored at regular intervals. The historic state of charge data can be queried over a plurality of intervals and a state of charge curve generated that is representative of a user's charging habits over time. The state of charge curve can be used to generate a rate of charge histogram and an acceleration of charge histogram. These can be used to predict when a user will charge next, and whether the energy storage device will have an amount of energy below a predetermined threshold amount before the next predicted charging time. A first device can determine when a second device typically charges and whether the energy storage device in the second device will have an amount of energy below the predetermined threshold amount before the next predicted charge time for the second device. The first device can generate an advice to charge notification to the user on either, or both, devices.

Smart advice to charge notification

A device implementing a system to execute machine learning models from memory includes at least one processor configured to receive a request to provide an input to one or more machine learning (ML) models arranged into a graph of connected layers, the one or more ML models stored in the first type of memory. The at least one processor is further configured to divide the graph of connected layers into a plurality of segments such that at least two of the plurality of segments concurrently fits within allocated space of the second type of memory. The at least one processor is further configured to cause the input to be processed through the first segment of the plurality of segments using the second type of memory while a second segment of the plurality of segments is concurrently loaded from the first type of memory into the second type of memory.

Execution of segmented machine learning models

The subject technology provides for dynamic task allocation for neural network models. The subject technology determines an operation performed at a node of a neural network model. The subject technology assigns an annotation to indicate whether the operation is better performed on a CPU or a GPU based at least in part on hardware capabilities of a target platform. The subject technology determines whether the neural network model includes a second layer. The subject technology, in response to determining that the neural network model includes a second layer, for each node of the second layer of the neural network model, determines a second operation performed at the node. Further the subject technology assigns a second annotation to indicate whether the second operation is better performed on the CPU or the GPU based at least in part on the hardware capabilities of the target platform.

Annotation override determination for a neural network

The subject technology provides receiving a neural network (NN) model to be executed on a target platform, the NN model including multiple layers that include operations and some of the operations being executable on multiple processors of the target platform. The subject technology further sorts the operations from the multiple layers in a particular order based at least in part on grouping the operations that are executable by a particular processor of the multiple processors. The subject technology determines, based at least in part on a cost of transferring the operations between the multiple processors, an assignment of one of the multiple processors for each of the sorted operations of each of the layers in a manner that minimizes a total cost of executing the operations. Further, for each layer of the NN model, the subject technology includes an annotation to indicate the processor assigned for each of the operations.

Compiling models for dedicated hardware

Disclosed herein is a technique for implementing a framework that enables application developers to enhance their applications with dynamic adjustment capabilities. Specifically, the framework, when utilized by an application on a mobile computing device that implements the framework, can enable the application to establish predictive models that can be used to identify meaningful behavioral patterns of an individual who uses the application. In turn, the predictive models can be used to preempt the individual's actions and provide an enhanced overall user experience. The framework is configured to interface with other software entities on the mobile computing device that conduct various analyses to identify appropriate times for the application to manage and update its predictive models. Such appropriate times can include, for example, identified periods of time where the individual is not operating the mobile computing device, as well as recognized conditions where power consumption is not a concern.

Methods and system for managing predictive models

Embodiments described herein provide for a non-transitory machine-readable medium storing instructions to cause one or more processors to receive, at a client device, a machine learning model from a server, detect a usage pattern for a content item, store an association between the content item and the detected usage pattern in local data, train the machine learning model using local data for the content item with the detected usage pattern to generate a trained machine learning model, generate an update for the machine learning model, privatize the update for the machine learning model, and transmit the privatized update for the machine learning model to the server.

User behavior model development with private federated learning

This application relates to features for a mobile device that allow the mobile device to assign utility values to applications and thereafter suggest applications for a user to execute. The suggested application can be derived from a list of applications that have been assigned a utility by software in the mobile device. The utility assignment of the individual applications from the list of applications can be performed based on the occurrence of an event, an environmental change, or a period of frequent application usage. A feedback mechanism is provided in some embodiments for more accurately assigning a utility to particular applications. The feedback mechanism can track what a user does during a period of suggestion for certain applications and thereafter modify the utility of applications based on what applications a user selects during the period of suggestion.

Application suggestion features

The subject technology runs a compiled neural network (NN) model on a particular processor with multiple priority queues for executing different processes, the compiled NN model being assigned to a particular priority queue, and the compiled NN model includes context switch instructions that were previously inserted into a neural network (NN) model from which the compiled NN model was compiled. The subject technology determines that a particular context switch instruction has been executed by the particular processor. The subject technology determines that a different process is waiting to be executed, the different process being assigned to a different priority queue and the different process being a higher priority process than the running compiled NN model. In response to executing the particular context switch instruction, the subject technology performs a context switch to the different process assigned to the different priority queue when the different process is waiting to be executed.

Embodiments described herein provide for a non-transitory machine-readable medium storing instructions to cause one or more processors to perform operations comprising receiving a machine learning model from a server at a client device, training the machine learning model using local data at the client device, generating an update for the machine learning model, the update including a weight vector that represents a difference between the received machine learning model and the trained machine learning model, privatizing the update for the machine learning model, and transmitting the privatized update for the machine learning model to the server.

Private federated learning with protection against reconstruction

The embodiments set forth techniques for implementing various 'prediction engines' that can be configured to provide different kinds of predictions within a mobile computing device. According to some embodiments, each prediction engine can assign itself as an 'expert' on one or more &#x201c;prediction categories&#x201d; within the mobile computing device. When a software application issues a request for a prediction for a particular category, and two or more prediction engines respond with their respective prediction(s), a &#x201c;prediction center&#x201d; can be configured to receive and process the predictions prior to responding to the request. Processing the predictions can involve removing duplicate information that exists across the predictions, sorting the predictions in accordance with confidence levels advertised by the prediction engines, and the like. In this manner, the prediction center can distill multiple predictions down into an optimized prediction and provide the optimized prediction to the software application.

Systems and methods for providing predictions to applications executing on a computing device

The subject technology transforms a machine learning model into a transformed machine learning model in accordance with a particular model specification when the machine learning model does not conform to the particular model specification, the particular model specification being compatible with an integrated development environment (IDE). The subject technology generates a code interface and code for the transformed machine learning model, the code interface including code statements in the object oriented programming language, the code statements corresponding to an object representing the transformed machine learning model. Further, the subject technology provides the generated code interface and the code for display in the IDE, the IDE enabling modifying of the generated code interface and the code.

Integration of learning models into a software development system

Dynamic task allocation for neural networks

In an exemplary process for remote execution of machine-learned models, one or more signals from a second electronic device is detected by a first electronic device. The second electronic device includes a machine-learned model associated with an application implemented on the first electronic device. Based on the one or more signals, a communication connection is established with the second electronic device and a proxy to the machine-learned model is generated. Input data is obtained via a sensor of the first electronic device. A representation of the input data is sent to the second electronic device via the proxy and the established communication connection. The representation of the input data is processed through the machine-learned model to generate an output. A result derived from the output is received via the communication connection and a representation of the result is outputted.

Remote execution of machine-learned models

Growing community of inventors

Santa Clara, CA, United States of America

Gaurav Kapoor