[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

US20210089918A1 - Systems and methods for cooperative machine learning - Google Patents

Systems and methods for cooperative machine learning Download PDF

Info

Publication number
US20210089918A1
US20210089918A1 US17/115,395 US202017115395A US2021089918A1 US 20210089918 A1 US20210089918 A1 US 20210089918A1 US 202017115395 A US202017115395 A US 202017115395A US 2021089918 A1 US2021089918 A1 US 2021089918A1
Authority
US
United States
Prior art keywords
machine learning
learning model
client
computing device
side machine
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Abandoned
Application number
US17/115,395
Inventor
John Rogers
Kevin Most
Matthew Zeiler
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Clarifai Inc
Original Assignee
Clarifai Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Clarifai Inc filed Critical Clarifai Inc
Priority to US17/115,395 priority Critical patent/US20210089918A1/en
Assigned to CLARIFAI, INC. reassignment CLARIFAI, INC. ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: MOST, KEVIN, ZEILER, MATTHEW, ROGERS, JOHN
Publication of US20210089918A1 publication Critical patent/US20210089918A1/en
Abandoned legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00Machine learning
    • G06N20/20Ensemble learning
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00Network arrangements or protocols for supporting network services or applications
    • H04L67/01Protocols
    • H04L67/42
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00Network arrangements or protocols for supporting network services or applications
    • H04L67/50Network services
    • H04L67/56Provisioning of proxy services
    • H04L67/59Providing operational support to end devices by off-loading in the network or by emulation, e.g. when they are unavailable

Definitions

  • This disclosure relates to systems and methods for cooperative machine learning, for example, across multiple client computing platforms, the cloud enabling off-line deep neural network operations on client computing platforms, etc.
  • Deep neural networks have traditionally been run on very fast computers with expensive graphics cards, due to their voracious appetite for computing power.
  • there are few implementations for mobile many of the models are too big, and the implementations are not easy to use.
  • Existing approaches use cloud machine learning APIs, but these suffer from very high latency.
  • Exemplary implementations disclosed herein may allow mobile applications to use deep learning to understand structured data like images, video, audio, and text.
  • exemplary implementations enable real-time applications of machine learning on mobile devices due to low latency facilitated by sharing of models between users.
  • Exemplary implementations also solve issues related to siloing. For example, mobile developers may want to use feedback from users to develop smart models. If there is no cloud for centralized learning, every model trained by a user will be only as good as that user has made it. With a central cloud, users can collaborate to train much smarter models, as well as sharing new ones between devices. This may enable a whole host of new applications and features for mobile developers.
  • the system may comprise one or more hardware processors configured by machine-readable instruction components.
  • the components may comprise a communications component, a cloud neural network component, a synchronization component, a cloud learning component, and an optimization component.
  • the communications component may be configured to facilitate communications between one or more client computing platforms and one or more servers.
  • the one or more client computing platforms may include a first client computing platform.
  • the first client computing platform may include a client-side machine learning model configured to facilitate deep neural network operations on structured data. The operations may be performed by a client application installed on the first client computing platform.
  • the client application may locally access the client-side machine learning model in order to perform the operations.
  • the cloud neural network component may include a cloud machine learning model configured to facilitate deep neural network operations on structured data.
  • the synchronization component may be configured to facilitate sharing of model states between the cloud machine learning model and the client-side machine learning model.
  • the cloud learning component may be configured to improve the cloud machine learning model based on usage of the application and user interactions with the first client computing platform.
  • the optimization component may be configured to determine workflows between one or more client computing platforms and one or more servers that chain multiple machine learning models.
  • the method may be performed by one or more hardware processors configured by machine-readable instructions.
  • the method may include facilitating, at one or more servers, communications between one or more client computing platforms and one or more servers.
  • the one or more client computing platforms may include a first client computing platform.
  • the first client computing platform may include a client-side machine learning model configured to facilitate deep neural network operations on structured data.
  • the operations may be performed by a client application installed on the first client computing platform.
  • the client application may locally access the client-side machine learning model in order to perform the operations.
  • the method may include facilitating, at one or more servers, deep neural network operations on structured data.
  • the method may include facilitating, at one or more servers, sharing of model states between the cloud machine learning model and the client-side machine learning model.
  • the method may include improving, at one or more servers, the cloud machine learning model based on usage of the application and user interactions with the first client computing platform.
  • the method may include determining, at one or more servers, workflows between one or more client computing platforms and one or more servers that chain multiple machine learning models.
  • FIG. 1 illustrates a system configured for cooperative machine learning across multiple client computing platforms and the cloud enabling off-line deep neural network operations on client computing platforms, in accordance with one or more implementations.
  • FIG. 2 illustrates a method for cooperative machine learning across multiple client computing platforms and the cloud enabling off-line deep neural network operations on client computing platforms, in accordance with one or more implementations.
  • FIG. 1 illustrates a system configured for cooperative machine learning across multiple client computing platforms and the cloud enabling off-line deep neural network operations on client computing platforms, in accordance with one or more implementations.
  • system 100 may include one or more server 102 .
  • the server(s) 102 and client computing platforms 104 may be configured to communicate according to a client/server architecture, a peer-to-peer architecture, and/or other architectures.
  • the users may access system 100 via client computing platform(s) 104 .
  • the client computing platform(s) 104 may be configured to execute machine-readable instructions 106 .
  • the machine-readable instructions 106 may include one or more of a client-side neural network component 108 , a client-side learning component 110 , and/or other components.
  • the client-side neural network component 108 may include a client-side machine learning model configured to facilitate deep neural network operations on structured data.
  • the structured data may include one or more of images, images within images, symbols, logos, objects, video, audio, text, and/or other structured data.
  • the deep neural network operations may relate to deep learning, deep structured learning, hierarchical learning, deep machine learning, and/or other types of machine learning.
  • a deep neural network operation may include a set of algorithms that attempt to model high level abstractions in data by using a deep graph with multiple processing layers, composed of multiple linear and/or non-linear transformations.
  • the deep neural network operations may be performed by a client application installed on a given client computing platform 104 .
  • the client application may include a software application designed to run on the given client computing platform 104 .
  • the client application may locally access the client-side machine learning model in order to perform the deep neural network operations. These operations may occur at very low latency because they are embedded on the client computing platform(s) 104 , as opposed to running in the cloud.
  • the client-side neural network component 108 may be embedded in the client application installed on the given client computing platform 104 .
  • the application may access the client-side neural network component 108 stored locally on the given client computing platform 104 .
  • the client-side neural network component 108 may include a code library enabling the deep neural network operations.
  • the client-side neural network component 108 may include multiple client-side learning models such that an arbitrary workflow can be applied to determine desired outputs.
  • the client-side learning component 110 may be configured to improve the client-side machine learning model based on one or more of new model states received from other locations in system 100 (described further below), usage of the application, user interactions with the given client computing platform 104 , one or more other client computing platforms 104 , and/or other information. According to some implementations, sharing of information beyond model states may facilitate learning for a distributed machine learning model from multiple client computing platform(s) 104 and/or server(s) 102 .
  • the client-side learning component 110 may be configured to improve the client-side machine learning model in real time or near-real time.
  • learning from this data may occur in a software developer kit (SDK).
  • SDK software developer kit
  • the SDK may provide a low latency and no-wireless-network-connection capability to recognize new things in the world.
  • learning may occur on an application programing interface (API) platform to benefit from data provided from multiple client computing platforms and/or other API platform users.
  • API application programing interface
  • the server(s) 102 may be configured to execute machine-readable instructions 112 .
  • the machine-readable instructions 112 may include one or more of a communications component 114 , a cloud neural network component 116 , a synchronization component 118 , a cloud learning component 120 , an optimization component 122 , and/or other components.
  • the communications component 114 may be configured to facilitate communications between client computing platform(s) 104 and server(s) 102 .
  • the cloud neural network component 116 may include a cloud machine learning model configured to facilitate deep neural network operations on structured data. These deep neural network operations may be the same as or similar to the ones facilitated by the client-side machine learning model described in connection with client-side neural network component 108 .
  • the synchronization component 118 may be configured to facilitate sharing of information between the cloud machine learning model and the client-side machine learning model. Such information may include model states of the cloud machine learning model and/or the client-side machine learning model.
  • the machine model states may include machine model parameters.
  • the sharing of model states occurs on a periodic basis. In some implementations, the sharing of model states occurs responsive to the client computing platform 104 coming online.
  • the synchronization component 118 may be configured to facilitate sharing of model states between different client-side machine learning models embodied at different client computing platform(s) 104 .
  • the synchronization component 118 may be configured to facilitate sharing of information beyond the model states between the cloud machine learning model and the client-side machine learning model.
  • Such information beyond the model states may include one or more of images, images within images, symbols, logos, objects, tags, videos, audio, geolocation, accelerometer data, metadata, and/or other information.
  • sharing of information beyond the model states may be subject to end user consent.
  • sharing of some information beyond the model states may be one-way sharing from the client computing platform 104 .
  • the cloud learning component 120 may be configured to improve the cloud machine learning model. Such improvements may be made based on one or more of new model states received from other locations in system 100 such as client computing platform(s) 104 , usage of the application, user interactions with client computing platform(s) 104 , and/or other information. In some implementations, the cloud learning component 120 may be configured to improve the cloud machine learning model in real time or near-real time.
  • advertisements may be pushed to individual client computing platforms 104 based on the cloud machine learning model and/or other information received from client computing platform(s) 104 .
  • the optimization component 122 may be configured to determine workflows between a given client computing platform 104 , one or more other client computing platforms 104 , and/or server(s) 102 that chain multiple machine learning models.
  • the workflows may be determined in an arbitrary graph.
  • the workflow determination may be based on one or more of availability of network connection, network bandwidth, latency, throughput, number of concepts recognized by the machine learning models, importance of low latency, importance of high accuracy, user's preferences, preferences associated with the given client computing platform 104 , and/or other information.
  • a set of cooperating client computing platforms 104 may be those of the company's users. For individual ones of those client computing platforms 104 , the company may obtain an understanding of the type of data their users encounter in the world.
  • her client computing platform 104 e.g., mobile phone
  • a user may be able to train a machine learning model of her pet inside her home. That specific machine learning model may be pushed to server(s) 102 .
  • the machine learning model may be deployed to run on the user's webcam to process live feeds of video in order to find any frames of video that contain her pet.
  • This data may be reported as all video frames sent back to server(s) 102 .
  • This data may be reported as select video frames depending on the recognition of the pet. This data may serve as analytical data (e.g., the fraction of frames where your pet was recognized).
  • Another non-limiting example may involve deploying recognition of the same concept to multiple devices for surveillance.
  • a machine learning model could be trained to recognize a concept, and then be deployed immediately so the data from each video frame could be fed back to a centralized dashboard via client computing platform(s) 104 and/or server(s) 102 to alert when the desired concepts are recognized.
  • server(s) 102 , client computing platform(s) 104 , and/or external resources 124 may be operatively linked via one or more electronic communication links.
  • electronic communication links may be established, at least in part, via a network such as the Internet and/or other networks. It will be appreciated that this is not intended to be limiting, and that the scope of this disclosure includes implementations in which server(s) 102 , client computing platform(s) 104 , and/or external resources 124 may be operatively linked via some other communication media.
  • a given client computing platform 104 may include electronic storage 126 , one or more processors 130 configured to execute machine-readable instructions, and/or other components.
  • the machine-readable instructions may be configured to enable a user associated with the given client computing platform 104 to interface with system 100 and/or external resources 124 , and/or provide other functionality attributed herein to client computing platform(s) 104 .
  • the given client computing platform 104 may include one or more of a desktop computer, a laptop computer, a handheld computer, a tablet computing platform, a NetBook, a Smartphone, a gaming console, and/or other computing platforms.
  • External resources 124 may include sources of information, hosts and/or providers of machine learning outside of system 100 , external entities participating with system 100 , and/or other resources. In some implementations, some or all of the functionality attributed herein to external resources 124 may be provided by resources included in system 100 .
  • Server(s) 102 may include electronic storage 128 , one or more processors 132 , and/or other components. Server(s) 102 may include communication lines, or ports to enable the exchange of information with a network and/or other computing platforms. Illustration of server(s) 102 in FIG. 1 is not intended to be limiting. Server(s) 102 may include a plurality of hardware, software, and/or firmware components operating together to provide the functionality attributed herein to server(s) 102 . For example, server(s) 102 may be implemented by a cloud of computing platforms operating together as server(s) 102 .
  • Electronic storage 126 and 128 may comprise non-transitory storage media that electronically stores information.
  • the electronic storage media of electronic storage 126 may include one or both of system storage that is provided integrally (i.e., substantially non-removable) with a given client computing platform 104 and/or removable storage that is removably connectable to the given client computing platform 104 via, for example, a port (e.g., a USB port, a firewire port, etc.) or a drive (e.g., a disk drive, etc.).
  • a port e.g., a USB port, a firewire port, etc.
  • a drive e.g., a disk drive, etc.
  • the electronic storage media of electronic storage 128 may include one or both of system storage that is provided integrally (i.e., substantially non-removable) with server(s) 102 and/or removable storage that is removably connectable to server(s) 102 via, for example, a port (e.g., a USB port, a firewire port, etc.) or a drive (e.g., a disk drive, etc.).
  • a port e.g., a USB port, a firewire port, etc.
  • a drive e.g., a disk drive, etc.
  • Electronic storage 126 and 128 may include one or more of optically readable storage media (e.g., optical disks, etc.), magnetically readable storage media (e.g., magnetic tape, magnetic hard drive, floppy drive, etc.), electrical charge-based storage media (e.g., EEPROM, RAM, etc.), solid-state storage media (e.g., flash drive, etc.), and/or other electronically readable storage media.
  • Electronic storage 128 may include one or more virtual storage resources (e.g., cloud storage, a virtual private network, and/or other virtual storage resources).
  • Electronic storage 126 and 128 may store software algorithms, information determined by processor(s) 126 and 128 , information received from server(s) 102 , information received from client computing platform(s) 104 , and/or other information that enables server(s) 102 and client puttinging platform(s) 104 to function as described herein.
  • Processor(s) 130 may be configured to provide information processing capabilities in client computing platform(s) 104 .
  • Processor(s) 132 may be configured to provide information processing capabilities in server(s) 102 .
  • processor(s) 130 and 132 may include one or more of a digital processor, an analog processor, a digital circuit designed to process information, an analog circuit designed to process information, a state machine, and/or other mechanisms for electronically processing information.
  • processor(s) 130 and 132 are shown in FIG. 1 as single entities, this is for illustrative purposes only.
  • processor(s) 130 and 132 may each include a plurality of processing units.
  • processor(s) 132 may represent processing functionality of a plurality of servers operating in coordination.
  • the processor(s) 130 and/or 132 may be configured to execute machine-readable instruction components 108 , 110 , 114 , 116 , 118 , 120 , 122 , and/or other machine-readable instruction components.
  • the processor(s) 130 and/or 132 may be configured to execute machine-readable instruction components 108 , 110 , 114 , 116 , 118 , 120 , 122 , and/or other machine-readable instruction components by software; hardware; firmware; some combination of software, hardware, and/or firmware; and/or other mechanisms for configuring processing capabilities on processor(s) 130 and/or 132 .
  • machine-readable instruction components 108 , 110 , 114 , 116 , 118 , 120 , and 122 are illustrated in FIG. 1 as being implemented within single processing units, in implementations in which processor(s) 130 and/or 132 include multiple processing units, one or more of machine-readable instruction components 108 , 110 , 114 , 116 , 118 , 120 , and/or 122 may be implemented remotely from the other components and/or subcomponents.
  • machine-readable instruction components 108 , 110 , 114 , 116 , 118 , 120 , and/or 122 described herein is for illustrative purposes, and is not intended to be limiting, as any of machine-readable instruction components 108 , 110 , 114 , 116 , 118 , 120 , and/or 122 may provide more or less functionality than is described.
  • machine-readable instruction components 108 , 110 , 114 , 116 , 118 , 120 , and/or 122 may be eliminated, and some or all of its functionality may be provided by other ones of machine-readable instruction components 108 , 110 , 114 , 116 , 118 , 120 , and/or 122 .
  • processor(s) 130 and/or 132 may be configured to execute one or more additional machine-readable instruction components that may perform some or all of the functionality attributed below to one of machine-readable instruction components 108 , 110 , 114 , 116 , 118 , 120 , and/or 122 .
  • FIG. 2 illustrates a method 200 for cooperative machine learning across multiple client computing platforms and the cloud enabling off-line deep neural network operations on client computing platforms, in accordance with one or more implementations.
  • the operations of method 200 presented below are intended to be illustrative. In some implementations, method 200 may be accomplished with one or more additional operations not described, and/or without one or more of the operations discussed. Additionally, the order in which the operations of method 200 are illustrated in FIG. 2 and described below is not intended to be limiting.
  • method 200 may be implemented in one or more processing devices (e.g., a digital processor, an analog processor, a digital circuit designed to process information, an analog circuit designed to process information, a state machine, and/or other mechanisms for electronically processing information).
  • the one or more processing devices may include one or more devices executing some or all of the operations of method 200 in response to instructions stored electronically on an electronic storage medium.
  • the one or more processing devices may include one or more devices configured through hardware, firmware, and/or software to be specifically designed for execution of one or more of the operations of method 200 .
  • a client computing platform 104 A may perform deep neural network operations on structured data.
  • the structured data may include one or more of images, images within images, symbols, logos, objects, video, audio, text, and/or other structured data obtained by or stored by the client computing platform 104 A.
  • the deep neural network operations may be performed by a client application installed on the client computing platform 104 A.
  • the client application may locally access the client-side machine learning model in order to perform the deep neural network operations.
  • Operation 202 may be performed by one or more hardware processors configured to execute a machine-readable instruction component that is the same as or similar to client-side neural network component 108 (as described in connection with FIG. 1 ), in accordance with one or more implementations.
  • the client computing platform 104 A may improve the client-side machine learning model based on one or more of new model states received from other locations in system 100 , usage of the application, user interactions with the given client computing platform 104 , and/or other information. Operation 204 may be performed by one or more hardware processors configured to execute a machine-readable instruction component that is the same as or similar to client-side learning component 110 (as described in connection with FIG. 1 ), in accordance with one or more implementations.
  • server(s) 102 may perform deep neural network operations on structured data. These deep neural network operations may be the same as or similar to the ones facilitated by the client-side machine learning model described in connection with client-side neural network component 108 . Operation 206 may be performed by one or more hardware processors configured to execute a machine-readable instruction component that is the same as or similar to cloud neural network component 116 (as described in connection with FIG. 1 ), in accordance with one or more implementations.
  • server(s) 102 may improve the cloud machine learning model. Such improvements may be made based on one or more of new model states received from other locations in system 100 , usage of the application, user interactions with client computing platform(s) 104 , and/or other information. Operation 208 may be performed by one or more hardware processors configured to execute a machine-readable instruction component that is the same as or similar to cloud learning component 120 (as described in connection with FIG. 1 ), in accordance with one or more implementations.
  • information may be shared between the cloud machine learning model and the client-side machine learning model.
  • Such information may include model states of the cloud machine learning model and/or the client-side machine learning model.
  • the machine model states may include machine model parameters.
  • the sharing may occur between client computing platform 104 A, server(s) 102 , and/or other client computing platform(s) 104 X.
  • Operations 210 A, 210 B, and/or 210 C may be performed by one or more hardware processors configured to execute a machine-readable instruction component that is the same as or similar to synchronization component 118 (as described in connection with FIG. 1 ), in accordance with one or more implementations.
  • information beyond the model states may be shared between the cloud machine learning model and the client-side machine learning model.
  • Such information beyond the model states may include one or more of images, images within images, symbols, logos, objects, tags, videos, audio, geolocation, accelerometer data, metadata, and/or other information.
  • sharing of information beyond the model states may be subject to end user consent. The sharing may occur between client computing platform 104 A, server(s) 102 , and/or other client computing platform(s) 104 X. In some implementations, sharing of some information beyond the model states may be one-way sharing from the client computing platform 104 .
  • Operations 212 A, 212 B, and/or 212 C may be performed by one or more hardware processors configured to execute a machine-readable instruction component that is the same as or similar to synchronization component 118 (as described in connection with FIG. 1 ), in accordance with one or more implementations.
  • Operations 214 and 216 illustrate that the learning processes of system 100 can be recursive.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Information Transfer Between Computers (AREA)
  • User Interface Of Digital Computer (AREA)

Abstract

In some embodiments, a given client computing platform may include a client-side machine learning model configured to facilitate deep neural network operations on structured data. The operations may be performed by a client application installed on the given client computing platform. The client application may locally access the client-side machine learning model in order to perform the operations. Deep neural network operations on structured data may be performed at one or more servers. Sharing of model states may be facilitated between the cloud machine learning model and the client-side machine learning model. The cloud machine learning model may be improved, at one or more servers, based on usage of the application and user interactions with the given client computing platform.

Description

    CROSS-REFERENCE TO RELATED APPLICATIONS
  • This application is a continuation of U.S. patent application Ser. No. 15/276,655, filed Sep. 26, 2016, the content of which is incorporated herein in its entirety by reference.
  • FIELD OF THE DISCLOSURE
  • This disclosure relates to systems and methods for cooperative machine learning, for example, across multiple client computing platforms, the cloud enabling off-line deep neural network operations on client computing platforms, etc.
  • BACKGROUND
  • Deep neural networks have traditionally been run on very fast computers with expensive graphics cards, due to their voracious appetite for computing power. Currently, however, it is difficult for mobile developers to use this technology. For example, there are few implementations for mobile, many of the models are too big, and the implementations are not easy to use. Existing approaches use cloud machine learning APIs, but these suffer from very high latency.
  • SUMMARY
  • Exemplary implementations disclosed herein may allow mobile applications to use deep learning to understand structured data like images, video, audio, and text. In particular, exemplary implementations enable real-time applications of machine learning on mobile devices due to low latency facilitated by sharing of models between users. Exemplary implementations also solve issues related to siloing. For example, mobile developers may want to use feedback from users to develop smart models. If there is no cloud for centralized learning, every model trained by a user will be only as good as that user has made it. With a central cloud, users can collaborate to train much smarter models, as well as sharing new ones between devices. This may enable a whole host of new applications and features for mobile developers.
  • Accordingly, one aspect of the disclosure relates to a system configured for cooperative machine learning across multiple client computing platforms and the cloud enabling off-line deep neural network operations on client computing platforms. The system may comprise one or more hardware processors configured by machine-readable instruction components. The components may comprise a communications component, a cloud neural network component, a synchronization component, a cloud learning component, and an optimization component. The communications component may be configured to facilitate communications between one or more client computing platforms and one or more servers. The one or more client computing platforms may include a first client computing platform. The first client computing platform may include a client-side machine learning model configured to facilitate deep neural network operations on structured data. The operations may be performed by a client application installed on the first client computing platform. The client application may locally access the client-side machine learning model in order to perform the operations. The cloud neural network component may include a cloud machine learning model configured to facilitate deep neural network operations on structured data. The synchronization component may be configured to facilitate sharing of model states between the cloud machine learning model and the client-side machine learning model. The cloud learning component may be configured to improve the cloud machine learning model based on usage of the application and user interactions with the first client computing platform. The optimization component may be configured to determine workflows between one or more client computing platforms and one or more servers that chain multiple machine learning models.
  • Another aspect of the disclosure relates to a method for cooperative machine learning across multiple client computing platforms and the cloud enabling off-line deep neural network operations on client computing platforms. The method may be performed by one or more hardware processors configured by machine-readable instructions. The method may include facilitating, at one or more servers, communications between one or more client computing platforms and one or more servers. The one or more client computing platforms may include a first client computing platform. The first client computing platform may include a client-side machine learning model configured to facilitate deep neural network operations on structured data. The operations may be performed by a client application installed on the first client computing platform. The client application may locally access the client-side machine learning model in order to perform the operations. The method may include facilitating, at one or more servers, deep neural network operations on structured data. The method may include facilitating, at one or more servers, sharing of model states between the cloud machine learning model and the client-side machine learning model. The method may include improving, at one or more servers, the cloud machine learning model based on usage of the application and user interactions with the first client computing platform. The method may include determining, at one or more servers, workflows between one or more client computing platforms and one or more servers that chain multiple machine learning models.
  • These and other features, and characteristics of the present technology, as well as the methods of operation and functions of the related elements of structure and the combination of parts and economies of manufacture, will become more apparent upon consideration of the following description and the appended claims with reference to the accompanying drawings, all of which form a part of this specification, wherein like reference numerals designate corresponding parts in the various figures. It is to be expressly understood, however, that the drawings are for the purpose of illustration and description only and are not intended as a definition of the limits of the invention. As used in the specification and in the claims, the singular form of “a”, “an”, and “the” include plural referents unless the context clearly dictates otherwise.
  • BRIEF DESCRIPTION OF THE DRAWINGS
  • FIG. 1 illustrates a system configured for cooperative machine learning across multiple client computing platforms and the cloud enabling off-line deep neural network operations on client computing platforms, in accordance with one or more implementations.
  • FIG. 2 illustrates a method for cooperative machine learning across multiple client computing platforms and the cloud enabling off-line deep neural network operations on client computing platforms, in accordance with one or more implementations.
  • DETAILED DESCRIPTION
  • FIG. 1 illustrates a system configured for cooperative machine learning across multiple client computing platforms and the cloud enabling off-line deep neural network operations on client computing platforms, in accordance with one or more implementations. In some implementations, system 100 may include one or more server 102. The server(s) 102 and client computing platforms 104 may be configured to communicate according to a client/server architecture, a peer-to-peer architecture, and/or other architectures. The users may access system 100 via client computing platform(s) 104.
  • The client computing platform(s) 104 may be configured to execute machine-readable instructions 106. The machine-readable instructions 106 may include one or more of a client-side neural network component 108, a client-side learning component 110, and/or other components.
  • The client-side neural network component 108 may include a client-side machine learning model configured to facilitate deep neural network operations on structured data. The structured data may include one or more of images, images within images, symbols, logos, objects, video, audio, text, and/or other structured data. The deep neural network operations may relate to deep learning, deep structured learning, hierarchical learning, deep machine learning, and/or other types of machine learning. In some implementations, a deep neural network operation may include a set of algorithms that attempt to model high level abstractions in data by using a deep graph with multiple processing layers, composed of multiple linear and/or non-linear transformations.
  • The deep neural network operations may be performed by a client application installed on a given client computing platform 104. The client application may include a software application designed to run on the given client computing platform 104. The client application may locally access the client-side machine learning model in order to perform the deep neural network operations. These operations may occur at very low latency because they are embedded on the client computing platform(s) 104, as opposed to running in the cloud.
  • In some implementations, the client-side neural network component 108 may be embedded in the client application installed on the given client computing platform 104. In some implementations, the application may access the client-side neural network component 108 stored locally on the given client computing platform 104. In some implementations, the client-side neural network component 108 may include a code library enabling the deep neural network operations. In some implementations, the client-side neural network component 108 may include multiple client-side learning models such that an arbitrary workflow can be applied to determine desired outputs.
  • The client-side learning component 110 may be configured to improve the client-side machine learning model based on one or more of new model states received from other locations in system 100 (described further below), usage of the application, user interactions with the given client computing platform 104, one or more other client computing platforms 104, and/or other information. According to some implementations, sharing of information beyond model states may facilitate learning for a distributed machine learning model from multiple client computing platform(s) 104 and/or server(s) 102.
  • In some implementations, the client-side learning component 110 may be configured to improve the client-side machine learning model in real time or near-real time. In some implementations, learning from this data may occur in a software developer kit (SDK). The SDK may provide a low latency and no-wireless-network-connection capability to recognize new things in the world. In some implementations, learning may occur on an application programing interface (API) platform to benefit from data provided from multiple client computing platforms and/or other API platform users.
  • The server(s) 102 may be configured to execute machine-readable instructions 112. The machine-readable instructions 112 may include one or more of a communications component 114, a cloud neural network component 116, a synchronization component 118, a cloud learning component 120, an optimization component 122, and/or other components.
  • The communications component 114 may be configured to facilitate communications between client computing platform(s) 104 and server(s) 102.
  • The cloud neural network component 116 may include a cloud machine learning model configured to facilitate deep neural network operations on structured data. These deep neural network operations may be the same as or similar to the ones facilitated by the client-side machine learning model described in connection with client-side neural network component 108.
  • The synchronization component 118 may be configured to facilitate sharing of information between the cloud machine learning model and the client-side machine learning model. Such information may include model states of the cloud machine learning model and/or the client-side machine learning model. The machine model states may include machine model parameters. In some implementations, the sharing of model states occurs on a periodic basis. In some implementations, the sharing of model states occurs responsive to the client computing platform 104 coming online. The synchronization component 118 may be configured to facilitate sharing of model states between different client-side machine learning models embodied at different client computing platform(s) 104.
  • The synchronization component 118 may be configured to facilitate sharing of information beyond the model states between the cloud machine learning model and the client-side machine learning model. Such information beyond the model states may include one or more of images, images within images, symbols, logos, objects, tags, videos, audio, geolocation, accelerometer data, metadata, and/or other information. In some implementations, sharing of information beyond the model states may be subject to end user consent. In some implementations, sharing of some information beyond the model states may be one-way sharing from the client computing platform 104.
  • The cloud learning component 120 may be configured to improve the cloud machine learning model. Such improvements may be made based on one or more of new model states received from other locations in system 100 such as client computing platform(s) 104, usage of the application, user interactions with client computing platform(s) 104, and/or other information. In some implementations, the cloud learning component 120 may be configured to improve the cloud machine learning model in real time or near-real time.
  • In some implementations, advertisements may be pushed to individual client computing platforms 104 based on the cloud machine learning model and/or other information received from client computing platform(s) 104.
  • The optimization component 122 may be configured to determine workflows between a given client computing platform 104, one or more other client computing platforms 104, and/or server(s) 102 that chain multiple machine learning models. In some implementations, the workflows may be determined in an arbitrary graph. In some implementations, the workflow determination may be based on one or more of availability of network connection, network bandwidth, latency, throughput, number of concepts recognized by the machine learning models, importance of low latency, importance of high accuracy, user's preferences, preferences associated with the given client computing platform 104, and/or other information.
  • By way of non-limiting example, for a company, a set of cooperating client computing platforms 104 may be those of the company's users. For individual ones of those client computing platforms 104, the company may obtain an understanding of the type of data their users encounter in the world.
  • As another illustrative example, using her client computing platform 104 (e.g., mobile phone), a user may be able to train a machine learning model of her pet inside her home. That specific machine learning model may be pushed to server(s) 102. The machine learning model may be deployed to run on the user's webcam to process live feeds of video in order to find any frames of video that contain her pet. This data may be reported as all video frames sent back to server(s) 102. This data may be reported as select video frames depending on the recognition of the pet. This data may serve as analytical data (e.g., the fraction of frames where your pet was recognized).
  • Another non-limiting example may involve deploying recognition of the same concept to multiple devices for surveillance. A machine learning model could be trained to recognize a concept, and then be deployed immediately so the data from each video frame could be fed back to a centralized dashboard via client computing platform(s) 104 and/or server(s) 102 to alert when the desired concepts are recognized.
  • In some implementations, server(s) 102, client computing platform(s) 104, and/or external resources 124 may be operatively linked via one or more electronic communication links. For example, such electronic communication links may be established, at least in part, via a network such as the Internet and/or other networks. It will be appreciated that this is not intended to be limiting, and that the scope of this disclosure includes implementations in which server(s) 102, client computing platform(s) 104, and/or external resources 124 may be operatively linked via some other communication media.
  • A given client computing platform 104 may include electronic storage 126, one or more processors 130 configured to execute machine-readable instructions, and/or other components. The machine-readable instructions may be configured to enable a user associated with the given client computing platform 104 to interface with system 100 and/or external resources 124, and/or provide other functionality attributed herein to client computing platform(s) 104. By way of non-limiting example, the given client computing platform 104 may include one or more of a desktop computer, a laptop computer, a handheld computer, a tablet computing platform, a NetBook, a Smartphone, a gaming console, and/or other computing platforms.
  • External resources 124 may include sources of information, hosts and/or providers of machine learning outside of system 100, external entities participating with system 100, and/or other resources. In some implementations, some or all of the functionality attributed herein to external resources 124 may be provided by resources included in system 100.
  • Server(s) 102 may include electronic storage 128, one or more processors 132, and/or other components. Server(s) 102 may include communication lines, or ports to enable the exchange of information with a network and/or other computing platforms. Illustration of server(s) 102 in FIG. 1 is not intended to be limiting. Server(s) 102 may include a plurality of hardware, software, and/or firmware components operating together to provide the functionality attributed herein to server(s) 102. For example, server(s) 102 may be implemented by a cloud of computing platforms operating together as server(s) 102.
  • Electronic storage 126 and 128 may comprise non-transitory storage media that electronically stores information. The electronic storage media of electronic storage 126 may include one or both of system storage that is provided integrally (i.e., substantially non-removable) with a given client computing platform 104 and/or removable storage that is removably connectable to the given client computing platform 104 via, for example, a port (e.g., a USB port, a firewire port, etc.) or a drive (e.g., a disk drive, etc.). The electronic storage media of electronic storage 128 may include one or both of system storage that is provided integrally (i.e., substantially non-removable) with server(s) 102 and/or removable storage that is removably connectable to server(s) 102 via, for example, a port (e.g., a USB port, a firewire port, etc.) or a drive (e.g., a disk drive, etc.). Electronic storage 126 and 128 may include one or more of optically readable storage media (e.g., optical disks, etc.), magnetically readable storage media (e.g., magnetic tape, magnetic hard drive, floppy drive, etc.), electrical charge-based storage media (e.g., EEPROM, RAM, etc.), solid-state storage media (e.g., flash drive, etc.), and/or other electronically readable storage media. Electronic storage 128 may include one or more virtual storage resources (e.g., cloud storage, a virtual private network, and/or other virtual storage resources). Electronic storage 126 and 128 may store software algorithms, information determined by processor(s) 126 and 128, information received from server(s) 102, information received from client computing platform(s) 104, and/or other information that enables server(s) 102 and client puttinging platform(s) 104 to function as described herein.
  • Processor(s) 130 may be configured to provide information processing capabilities in client computing platform(s) 104. Processor(s) 132 may be configured to provide information processing capabilities in server(s) 102. As such, processor(s) 130 and 132 may include one or more of a digital processor, an analog processor, a digital circuit designed to process information, an analog circuit designed to process information, a state machine, and/or other mechanisms for electronically processing information. Although processor(s) 130 and 132 are shown in FIG. 1 as single entities, this is for illustrative purposes only. In some implementations, processor(s) 130 and 132 may each include a plurality of processing units. For example, processor(s) 132 may represent processing functionality of a plurality of servers operating in coordination. The processor(s) 130 and/or 132 may be configured to execute machine- readable instruction components 108, 110, 114, 116, 118, 120, 122, and/or other machine-readable instruction components. The processor(s) 130 and/or 132 may be configured to execute machine- readable instruction components 108, 110, 114, 116, 118, 120, 122, and/or other machine-readable instruction components by software; hardware; firmware; some combination of software, hardware, and/or firmware; and/or other mechanisms for configuring processing capabilities on processor(s) 130 and/or 132.
  • It should be appreciated that although machine- readable instruction components 108, 110, 114, 116, 118, 120, and 122 are illustrated in FIG. 1 as being implemented within single processing units, in implementations in which processor(s) 130 and/or 132 include multiple processing units, one or more of machine- readable instruction components 108, 110, 114, 116, 118, 120, and/or 122 may be implemented remotely from the other components and/or subcomponents. The description of the functionality provided by the different machine- readable instruction components 108, 110, 114, 116, 118, 120, and/or 122 described herein is for illustrative purposes, and is not intended to be limiting, as any of machine- readable instruction components 108, 110, 114, 116, 118, 120, and/or 122 may provide more or less functionality than is described. For example, one or more of machine- readable instruction components 108, 110, 114, 116, 118, 120, and/or 122 may be eliminated, and some or all of its functionality may be provided by other ones of machine- readable instruction components 108, 110, 114, 116, 118, 120, and/or 122. As another example, processor(s) 130 and/or 132 may be configured to execute one or more additional machine-readable instruction components that may perform some or all of the functionality attributed below to one of machine- readable instruction components 108, 110, 114, 116, 118, 120, and/or 122.
  • FIG. 2 illustrates a method 200 for cooperative machine learning across multiple client computing platforms and the cloud enabling off-line deep neural network operations on client computing platforms, in accordance with one or more implementations. The operations of method 200 presented below are intended to be illustrative. In some implementations, method 200 may be accomplished with one or more additional operations not described, and/or without one or more of the operations discussed. Additionally, the order in which the operations of method 200 are illustrated in FIG. 2 and described below is not intended to be limiting.
  • In some implementations, method 200 may be implemented in one or more processing devices (e.g., a digital processor, an analog processor, a digital circuit designed to process information, an analog circuit designed to process information, a state machine, and/or other mechanisms for electronically processing information). The one or more processing devices may include one or more devices executing some or all of the operations of method 200 in response to instructions stored electronically on an electronic storage medium. The one or more processing devices may include one or more devices configured through hardware, firmware, and/or software to be specifically designed for execution of one or more of the operations of method 200.
  • At an operation 202, a client computing platform 104A may perform deep neural network operations on structured data. The structured data may include one or more of images, images within images, symbols, logos, objects, video, audio, text, and/or other structured data obtained by or stored by the client computing platform 104A. The deep neural network operations may be performed by a client application installed on the client computing platform 104A. The client application may locally access the client-side machine learning model in order to perform the deep neural network operations. Operation 202 may be performed by one or more hardware processors configured to execute a machine-readable instruction component that is the same as or similar to client-side neural network component 108 (as described in connection with FIG. 1), in accordance with one or more implementations.
  • At an operation 204, the client computing platform 104A may improve the client-side machine learning model based on one or more of new model states received from other locations in system 100, usage of the application, user interactions with the given client computing platform 104, and/or other information. Operation 204 may be performed by one or more hardware processors configured to execute a machine-readable instruction component that is the same as or similar to client-side learning component 110 (as described in connection with FIG. 1), in accordance with one or more implementations.
  • At an operation 206, server(s) 102 may perform deep neural network operations on structured data. These deep neural network operations may be the same as or similar to the ones facilitated by the client-side machine learning model described in connection with client-side neural network component 108. Operation 206 may be performed by one or more hardware processors configured to execute a machine-readable instruction component that is the same as or similar to cloud neural network component 116 (as described in connection with FIG. 1), in accordance with one or more implementations.
  • At an operation 208, server(s) 102 may improve the cloud machine learning model. Such improvements may be made based on one or more of new model states received from other locations in system 100, usage of the application, user interactions with client computing platform(s) 104, and/or other information. Operation 208 may be performed by one or more hardware processors configured to execute a machine-readable instruction component that is the same as or similar to cloud learning component 120 (as described in connection with FIG. 1), in accordance with one or more implementations.
  • At operations 210A, 210B, and/or 210C, information may be shared between the cloud machine learning model and the client-side machine learning model. Such information may include model states of the cloud machine learning model and/or the client-side machine learning model. The machine model states may include machine model parameters. The sharing may occur between client computing platform 104A, server(s) 102, and/or other client computing platform(s) 104X. Operations 210A, 210B, and/or 210C may be performed by one or more hardware processors configured to execute a machine-readable instruction component that is the same as or similar to synchronization component 118 (as described in connection with FIG. 1), in accordance with one or more implementations.
  • At operations 212A, 212B, and/or 212C, information beyond the model states may be shared between the cloud machine learning model and the client-side machine learning model. Such information beyond the model states may include one or more of images, images within images, symbols, logos, objects, tags, videos, audio, geolocation, accelerometer data, metadata, and/or other information. In some implementations, sharing of information beyond the model states may be subject to end user consent. The sharing may occur between client computing platform 104A, server(s) 102, and/or other client computing platform(s) 104X. In some implementations, sharing of some information beyond the model states may be one-way sharing from the client computing platform 104. Operations 212A, 212B, and/or 212C may be performed by one or more hardware processors configured to execute a machine-readable instruction component that is the same as or similar to synchronization component 118 (as described in connection with FIG. 1), in accordance with one or more implementations.
  • Operations 214 and 216 illustrate that the learning processes of system 100 can be recursive.
  • Although the present technology has been described in detail for the purpose of illustration based on what is currently considered to be the most practical and preferred implementations, it is to be understood that such detail is solely for that purpose and that the technology is not limited to the disclosed implementations, but, on the contrary, is intended to cover modifications and equivalent arrangements that are within the spirit and scope of the appended claims. For example, it is to be understood that the present technology contemplates that, to the extent possible, one or more features of any implementation can be combined with one or more features of any other implementation.

Claims (20)

What is claimed is:
1. One or more non-transitory computer-readable media storing instructions that, when executed by one or more processors, cause operations comprising:
identifying a client-side machine learning model hosted on a user computing device, the client-side machine learning model being trained based on a training dataset;
obtaining model parameters of another machine learning model hosted on another computing device, the other machine learning model being trained at the other computing device based on another training dataset, the model parameters of the other machine learning model being determined via the training of the other machine learning model at the other computing device; and
when the user computing device is online, updating the client-side machine learning model based on the model parameters of the other machine learning model.
2. The media of claim 1, the operations further comprising:
obtaining, from the user computing device, model parameters of the client-side machine learning model, the model parameters of the client-side machine learning model being determined via the training of the client-side machine learning model at the user computing device; and
providing, to the other computing device, the model parameters of the client-side machine learning model to update the other machine learning model based on the model parameters of the client-side machine learning model, the model parameters of the client-side machine learning model indicating a model state of the client-side machine learning model.
3. The media of claim 1, the operations further comprising:
accessing the client-side machine learning model while the client-side machine learning model is hosted on the user computing device; and
obtaining the training dataset and providing the training dataset as input for the client-side machine learning model to train the client-side machine learning model, the client-side machine learning model being trained based on the training dataset when the user computing device is offline from a computer system that comprises the one or more processors.
4. The media of claim 1, wherein the user computing device is a physical computing device remote from a computer system that comprises the one or more processors.
5. The media of claim 1, wherein the user computing device is a mobile computing device.
6. The media of claim 1, wherein the user computing device and the other computing device are physical computing devices remote from a computer system that comprises the one or more processors.
7. The media of claim 1, wherein the user computing device and the other computing device are mobile computing devices.
8. The media of claim 1, wherein the training dataset and the other training dataset comprise images, symbols, logos, videos, audio, text, geolocation, accelerometer data, or metadata.
9. A method comprising:
identifying, by one or more processors, a client-side machine learning model hosted on a user computing device, the client-side machine learning model being trained based on a training dataset;
obtaining, by the one or more processors, model parameters of another machine learning model hosted on another computing device, the other machine learning model being trained at the other computing device based on another training dataset, the model parameters of the other machine learning model being determined via the training of the other machine learning model at the other computing device; and
updating, by the one or more processors, the client-side machine learning model based on the model parameters of the other machine learning model.
10. The method of claim 9, further comprising:
obtaining, from the user computing device, model parameters of the client-side machine learning model, the model parameters of the client-side machine learning model being determined via the training of the client-side machine learning model at the user computing device; and
providing, to the other computing device, the model parameters of the client-side machine learning model to update the other machine learning model based on the model parameters of the client-side machine learning model, the model parameters of the client-side machine learning model indicating a model state of the client-side machine learning model.
11. The method of claim 9, further comprising:
accessing the client-side machine learning model while the client-side machine learning model is hosted on the user computing device; and
obtaining the training dataset and providing the training dataset as input for the client-side machine learning model to train the client-side machine learning model, the client-side machine learning model being trained based on the training dataset when the user computing device is offline from a computer system that comprises the one or more processors.
12. The method of claim 9, wherein the user computing device is a physical computing device remote from a computer system that comprises the one or more processors.
13. The method of claim 9, wherein the user computing device is a mobile computing device.
14. The method of claim 9, wherein the user computing device and the other computing device are physical computing devices remote from a computer system that comprises the one or more processors.
15. The method of claim 9, wherein the user computing device and the other computing device are mobile computing devices.
16. The method of claim 9, wherein the training dataset and the other training dataset comprise images, symbols, logos, videos, audio, text, geolocation, accelerometer data, or metadata.
17. A computer system comprising one or more processors programmed with instructions that, when executed by the one or more processors, cause operations comprising:
identifying a client-side machine learning model hosted on a user computing device, the client-side machine learning model being trained based on a training dataset;
obtaining model parameters of another machine learning model hosted on another computing device, the other machine learning model being trained at the other computing device based on another training dataset, the model parameters of the other machine learning model being determined via the training of the other machine learning model at the other computing device; and
updating the client-side machine learning model based on the model parameters of the other machine learning model.
18. The computer system of claim 17, the operations further comprising:
obtaining, from the user computing device, model parameters of the client-side machine learning model, the model parameters of the client-side machine learning model being determined via the training of the client-side machine learning model at the user computing device; and
providing, to the other computing device, the model parameters of the client-side machine learning model to update the other machine learning model based on the model parameters of the client-side machine learning model, the model parameters of the client-side machine learning model indicating a model state of the client-side machine learning model.
19. The computer system of claim 17, the operations further comprising:
accessing the client-side machine learning model while the client-side machine learning model is hosted on the user computing device; and
obtaining the training dataset and providing the training dataset as input for the client-side machine learning model to train the client-side machine learning model, the client-side machine learning model being trained based on the training dataset when the user computing device is offline from the computer system.
20. The computer system of claim 17, wherein the user computing device is a physical computing device remote from the computer system.
US17/115,395 2016-09-26 2020-12-08 Systems and methods for cooperative machine learning Abandoned US20210089918A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/115,395 US20210089918A1 (en) 2016-09-26 2020-12-08 Systems and methods for cooperative machine learning

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US15/276,655 US10867241B1 (en) 2016-09-26 2016-09-26 Systems and methods for cooperative machine learning across multiple client computing platforms and the cloud enabling off-line deep neural network operations on client computing platforms
US17/115,395 US20210089918A1 (en) 2016-09-26 2020-12-08 Systems and methods for cooperative machine learning

Related Parent Applications (1)

Application Number Title Priority Date Filing Date
US15/276,655 Continuation US10867241B1 (en) 2016-09-26 2016-09-26 Systems and methods for cooperative machine learning across multiple client computing platforms and the cloud enabling off-line deep neural network operations on client computing platforms

Publications (1)

Publication Number Publication Date
US20210089918A1 true US20210089918A1 (en) 2021-03-25

Family

ID=73746618

Family Applications (2)

Application Number Title Priority Date Filing Date
US15/276,655 Active 2038-05-21 US10867241B1 (en) 2016-09-26 2016-09-26 Systems and methods for cooperative machine learning across multiple client computing platforms and the cloud enabling off-line deep neural network operations on client computing platforms
US17/115,395 Abandoned US20210089918A1 (en) 2016-09-26 2020-12-08 Systems and methods for cooperative machine learning

Family Applications Before (1)

Application Number Title Priority Date Filing Date
US15/276,655 Active 2038-05-21 US10867241B1 (en) 2016-09-26 2016-09-26 Systems and methods for cooperative machine learning across multiple client computing platforms and the cloud enabling off-line deep neural network operations on client computing platforms

Country Status (1)

Country Link
US (2) US10867241B1 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2024134433A1 (en) * 2022-12-22 2024-06-27 Lumana Inc. Hybrid machine learning architecture for visual content processing and uses thereof

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11537262B1 (en) 2015-07-21 2022-12-27 Monotype Imaging Inc. Using attributes for font recommendations
US11334750B2 (en) * 2017-09-07 2022-05-17 Monotype Imaging Inc. Using attributes for predicting imagery performance
US11657602B2 (en) 2017-10-30 2023-05-23 Monotype Imaging Inc. Font identification from imagery
CN113741863A (en) * 2021-07-29 2021-12-03 南方电网深圳数字电网研究院有限公司 Application program generation method based on algorithm model, electronic device and storage medium

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110320520A1 (en) * 2010-06-23 2011-12-29 Microsoft Corporation Dynamic partitioning of applications between clients and servers
US20120191631A1 (en) * 2011-01-26 2012-07-26 Google Inc. Dynamic Predictive Modeling Platform
US20150170053A1 (en) * 2013-12-13 2015-06-18 Microsoft Corporation Personalized machine learning models
US20160350658A1 (en) * 2015-06-01 2016-12-01 Microsoft Technology Licensing, Llc Viewport-based implicit feedback
US20170366562A1 (en) * 2016-06-15 2017-12-21 Trustlook Inc. On-Device Maliciousness Categorization of Application Programs for Mobile Devices
US20180083898A1 (en) * 2016-09-20 2018-03-22 Google Llc Suggested responses based on message stickers
US9967226B2 (en) * 2014-06-30 2018-05-08 Microsoft Technology Licensing, Llc Personalized delivery time optimization

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150081543A1 (en) * 2013-09-16 2015-03-19 International Business Machines Corporation Analytics driven assessment of transactional risk daily limit exceptions

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110320520A1 (en) * 2010-06-23 2011-12-29 Microsoft Corporation Dynamic partitioning of applications between clients and servers
US20120191631A1 (en) * 2011-01-26 2012-07-26 Google Inc. Dynamic Predictive Modeling Platform
US20150170053A1 (en) * 2013-12-13 2015-06-18 Microsoft Corporation Personalized machine learning models
US9967226B2 (en) * 2014-06-30 2018-05-08 Microsoft Technology Licensing, Llc Personalized delivery time optimization
US20160350658A1 (en) * 2015-06-01 2016-12-01 Microsoft Technology Licensing, Llc Viewport-based implicit feedback
US20170366562A1 (en) * 2016-06-15 2017-12-21 Trustlook Inc. On-Device Maliciousness Categorization of Application Programs for Mobile Devices
US20180083898A1 (en) * 2016-09-20 2018-03-22 Google Llc Suggested responses based on message stickers

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2024134433A1 (en) * 2022-12-22 2024-06-27 Lumana Inc. Hybrid machine learning architecture for visual content processing and uses thereof

Also Published As

Publication number Publication date
US10867241B1 (en) 2020-12-15

Similar Documents

Publication Publication Date Title
US20210089918A1 (en) Systems and methods for cooperative machine learning
US20210342745A1 (en) Artificial intelligence model and data collection/development platform
US11681943B2 (en) Artificial intelligence development via user-selectable/connectable model representations
US10728345B2 (en) Field service management mobile offline synchronization
US10083054B2 (en) Application-based computing resource management
KR20200094207A (en) Methods and systems for generating personalized emoticons and lip syncing videos based on face recognition
US10616531B2 (en) Video feeds in collaboration environments
US10652075B2 (en) Systems and methods for selecting content items and generating multimedia content
US10798399B1 (en) Adaptive video compression
US10268549B2 (en) Heuristic process for inferring resource dependencies for recovery planning
US20180276478A1 (en) Determining Most Representative Still Image of a Video for Specific User
US10318639B2 (en) Intelligent action recommendation
US11159631B2 (en) Integration of social interactions into media sharing
US10348663B2 (en) Integration of social interactions into media sharing
US11574019B2 (en) Prediction integration for data management platforms
CN111492394A (en) Attendee engagement determination systems and methods
US20210117775A1 (en) Automated selection of unannotated data for annotation based on features generated during training
US11061982B2 (en) Social media tag suggestion based on product recognition
US10423821B2 (en) Automated profile image generation based on scheduled video conferences
US10706074B2 (en) Embeddings with multiple relationships
US9298744B2 (en) Method and apparatus for ordering images in an image set based on social interactions and viewer preferences
US10140518B1 (en) Content information auditing service
US20180061258A1 (en) Data driven feature discovery
US10924586B2 (en) Aggregating virtual reality (VR) sessions
US20230144585A1 (en) Machine learning model change detection and versioning

Legal Events

Date Code Title Description
AS Assignment

Owner name: CLARIFAI, INC., NEW YORK

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:ROGERS, JOHN;MOST, KEVIN;ZEILER, MATTHEW;SIGNING DATES FROM 20150526 TO 20181224;REEL/FRAME:054581/0554

STPP Information on status: patent application and granting procedure in general

Free format text: APPLICATION DISPATCHED FROM PREEXAM, NOT YET DOCKETED

STPP Information on status: patent application and granting procedure in general

Free format text: DOCKETED NEW CASE - READY FOR EXAMINATION

STPP Information on status: patent application and granting procedure in general

Free format text: NON FINAL ACTION MAILED

STPP Information on status: patent application and granting procedure in general

Free format text: RESPONSE TO NON-FINAL OFFICE ACTION ENTERED AND FORWARDED TO EXAMINER

STPP Information on status: patent application and granting procedure in general

Free format text: FINAL REJECTION MAILED

STCB Information on status: application discontinuation

Free format text: ABANDONED -- FAILURE TO RESPOND TO AN OFFICE ACTION