# US 2018/0025268 A1 · Configurable machine learning assemblies for autonomous operation in personal devices > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2018/0025268 A1 - Document ID: US20180025268A1 - Status: Published application - Kind: A1 - Application number: 15/396,267 - Filed: 2016-12-30 - Published: 2018-01-25 - Earliest claimed: 2016-07-21 - Assignees: Tessera Advanced Technologies, Inc. - Applicants: Tessera Advanced Technologies, Inc. - Inventors: Steven L. TEIG; Belgacem HABA; Ilyas MOHAMMED - Web page: https://patentgrep.com/patent/US20180025268A1 ## CPC - G06N 3/063: using electronic means - G06N 20/00: Machine learning - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract Configurable machine learning assemblies for autonomous operation in personal devices are provided. Example systems implement machine learning based on neural networks that draw low power for use in smart phones, watches, drones, automobiles, and medical devices. The onboard machine learning assemblies can be powered by batteries, and once onboard a small personal device, can learn to perform object recognition and autonomous decision-making without access to outside resources. The assemblies can be small or even nano-scale, and may draw less than one watt of power on average. An assembly can be configured from pluggable, interchangeable modules that have compatible ports for interconnecting and integrating functionally dissimilar sensor systems. A core module contains a machine learning kernel, and multiple cores can be connected together to expand the neural network. An example machine learning assembly auto-detects sensors and peripherals, and extends a network or bus to all connected components. ## Claims ### Claim 1 A system, comprising: multiple modules capable of being coupled with each other in different configurations; at least one core module of the multiple modules comprising a machine learning core based on a neural network; each module of the multiple modules having at least one instance of a same interface for communicatively coupling with at least one other module of the multiple modules; and the multiple modules reconfigurably coupled with each other to form an autonomous machine learning device. ### Claim 2 (depends on 1) The system of claim 1, wherein the multiple modules further comprise one or more peripheral modules capable of communicatively coupling with the at least one core module or with another peripheral module, wherein at least one peripheral module comprises a sensor module for providing data to the at least one core module containing the machine learning core. ### Claim 3 (depends on 2) The system of claim 2, further comprising different types of sensor modules, each different type of sensor module comprising a different type of sensor; and wherein the different types of sensor modules are attachably compatible with the autonomous machine learning device and detachably and reattachably interchangeable with each other to make different configurations of the autonomous machine learning device. ### Claim 4 (depends on 2) The system of claim 2, further comprising a configuration engine associated with the at least one core module to auto-detect each peripheral module being joined to the autonomous machine learning device. ### Claim 5 (depends on 2) The system of claim 2, further comprising an extensible local network controller or an extensible local bus controller to dynamically extend a local communication and control network or bus across the at least one core module and the one or more peripheral modules being attached, detached, or reattached to the autonomous machine learning device. ### Claim 6 (depends on 2) The system of claim 2, further comprising a peripheral modules tracker to track a performance and changes in a configuration of each of the one or more peripheral modules. ### Claim 7 (depends on 1) The system of claim 1, wherein multiple instances of the core module are communicatively coupled with each other via respective instances of the same interface to expand the neural network of the machine learning core in the autonomous machine learning device. ### Claim 8 (depends on 1) The system of claim 1, further comprising multiple instances of the machine learning core coupled with each other within the same module or via respective instances of the same interface between modules; and wherein the multiple instances of the machine learning core are coupled with each other in a series configuration, in a parallel processing configuration, in a tree configuration, or in a cluster configuration in the autonomous machine learning device. ### Claim 9 (depends on 1) The system of claim 1, wherein the autonomous machine learning device is scaled to a size selected from the group consisting of a nanomachine size comprising the multiple modules interconnected into a nanometer scale autonomous machine learning device, and a mobile device size comprising the interconnected multiple modules, each module having a size of approximately 6-20 millimeters. ### Claim 10 (depends on 1) The system of claim 1, wherein nodes of the neural network further comprise processing nodes with associated activation functions; and the at least one core module further comprising a neural network controller to apply specific activation functions to specific respective processing nodes of the neural network and to differentially weight data input to the respective processing nodes of the neural network. ### Claim 11 (depends on 1) The system of claim 1, wherein the autonomous machine-learning device receives information from one or more sensor modules; and further comprising a logic structure to self-improve a decision or conclusion based on subsequent iterations of a neural network function on the information from the one or more sensor modules, the information processed through the neural network in light of other previously processed or input information. ### Claim 12 (depends on 1) The system of claim 1, further comprising a first type of the same interface with male connection features and a second type of the same interface with female connections features; and further comprising an interconnection component releasably and reusably connecting a first module having the first type of the same interface with a second module having the second type of the same interface. ### Claim 13 (depends on 1) The system of claim 1, wherein the same interface comprises a coupling selected from the group consisting of a wireless communicative coupling between a first module and a second module, a physical and electrical coupling between the first module and the second module, a magnetic and electrical coupling between the first module and the second module, and a reversible physical and electrical coupling between the first module and the second module. ### Claim 14 (depends on 1) The system of claim 1, wherein the same interface further comprises a reversible physical interface and a reversible electrical interface with symmetrical physical features and symmetrical electrical features for coupling a first module and a second module in multiple orientations. ### Claim 15 (depends on 14) The system of claim 14, wherein the same interface comprises a port, wherein the port comprises a single channel port or a multiple channel port, and each module comprises one or multiple of the ports. ### Claim 16 (depends on 1) The system of claim 1, further comprising a device controlled or informed by the autonomous machine learning device; and wherein the autonomous machine learning device is contained within the device being controlled or informed and operates in isolation from communication or data transfer outside the device being controlled or informed by the autonomous machine learning device. ### Claim 17 (depends on 1) The system of claim 1, wherein the autonomous machine learning device comprises an imaging system selected from the group consisting of a microelectronic assembly for recognizing objects, a microelectronic assembly for recognizing faces, a microelectronic assembly for classifying objects in images, a microelectronic assembly for classifying images, and a microelectronic assembly for recognizing moving pedestrians. ### Claim 18 (depends on 1) The system of claim 1, wherein the autonomous machine learning device is incorporated into a hosting device selected from the group consisting of a smart phone, a tablet computing device, a drone, a portable camera, an automobile, an implantable medical device, a toy, a smart home, a smart factory, a smart city, and a security system. ### Claim 19 (depends on 1) The system of claim 1, wherein the autonomous machine learning device autonomously generates an executive output selected from the group consisting of steering a vehicle, navigating a drone or air vehicle, identifying a person, and authenticating a user. ### Claim 20 An apparatus, comprising: a sensor; a field programmable gate array device configured as a neural network to process a neural network function on information received from the sensor; and the neural network configured to generate a decision based on the information processed through the neural network in light of other information previously processed through the neural network. ### Claim 21 (depends on 20) The field programmable gate array device of claim 20, wherein the field programmable gate array device is physically and electrically coupled to the sensor. ### Claim 22 (depends on 21) The field programmable gate array device of claim 21, wherein the device self improves the decision or conclusion through processing subsequent iterations of the neural network function. ### Claim 23 A method comprising: generating first data based on a physical property sensed by a sensor of a portable personal device; providing the first data to a machine learning component coupled to the sensor of the portable personal device; applying a neural network function of the machine learning component to the first data to generate a signal; and executing a machine action or a device action of the portable personal device based on the signal. ### Claim 24 (depends on 23) The method of claim 23, further comprising adapting or refining the machine action or the device action based on applying the neural network function to second data from the sensor to provide a second signal. ### Claim 25 (depends on 24) The method of claim 24, wherein the machine learning component identifies a pattern based on the first data and the second data.