Inter-nodal sharing of generative models

Abstract

A machine facilitates inter-nodal sharing of a generative model by accessing a machine learning model that includes a plurality of nodes. Each node is configured to produce future outputs from future inputs based on a generative model of that node. Each node is also configured to update its generative model based on past feedback received by that node in response to past outputs of that node. The machine associates a first node with a second node based on a comparison of first updates of a first generative model of the first node to second updates of a second generative model of the second node. The machine determines a replacement generative model to replace the second generative model of the second node. The machine replaces the second generative model of the second node with the determined replacement generative model, thus configuring the second node to use the replacement generative model.

Description

INTER-NODAL SHARING OF GENERATIVE MODELSTECHNICAL FIELD

[0001] The subject matter disclosed herein generally relates to the technical field of special-purpose machines that facilitate artificial intelligence (Al), including software-configured computerized variants of such specialpurpose machines and improvements to such variants, and to the technologies by which such special-purpose machines become improved compared to other special-purpose machines that facilitate Al. Specifically, the present disclosure addresses systems and methods to facilitate inter-nodal sharing of generative models.BACKGROUND

[0002] Within the field of Al, machine learning models implement one or more algorithms or other techniques for performance on computer hardware and are configured to learn from experiences processing data and make decisions (e.g., predictions, inferences, or categorizations) without explicit programming. Machine learning models may utilize one or more statistical methods, mathematical optimization, pattern recognition techniques, or any suitable combination thereof, to identify patterns or other relationships within data. Machine learning models are now widely used across various domains, such as image recognition, speech recognition, natural language processing, classifiers, recommendation generators, and anomaly detectors. In traditional machine learning models, the model is trained using historical information that is labeled with one or more features, and after training, the trained machine learning model is provided with new unlabeled input data to generate inferences about the new data.BRIEF DESCRIPTION OF THE DRAWINGS

[0003] Some embodiments are illustrated by way of example and not limitation in the figures of the accompanying drawings.

[0004] FIG. l is a network diagram illustrating a network environment suitable for inter-nodal sharing of generative models, according to some example embodiments.

[0005] FIG. 2 is a block diagram illustrating components of an Al machine suitable for inter-nodal sharing of generative models, according to some example embodiments.

[0006] FIG. 3 is a conceptual diagram illustrating evolution of generative models (e.g., along respective evolutionary paths) in two Al nodes with similar initial sequences of updates to their generative models, according to some example embodiments.

[0007] FIG. 4 is a conceptual diagram illustrating evolution of generative models (e.g., along respective evolutionary paths) in two Al nodes with similar most recent sequences of updates to their generative models, according to some example embodiments.

[0008] FIGS. 5-11 are flowcharts illustrating operations of the Al machine in performing a method of inter-nodal sharing of a generative model, according to some example embodiments.

[0009] FIG. 12 is a block diagram illustrating components of a machine, according to some example embodiments, able to read instructions from a machine-readable medium and perform any one or more of the methodologies discussed herein.DETAILED DESCRIPTION

[0010] Example methods (e.g., algorithms) facilitate inter-nodal sharing of generative models (e.g., sharing a generative model between or among nodes, such as Al nodes), and example systems (e.g., special-purpose machines configured by special-purpose software) are configured to facilitate inter-nodal sharing of generative models. Examples merely typify possible variations.Unless explicitly stated otherwise, structures (e.g., structural components, such as modules) are optional and may be combined or subdivided, and operations (e.g., in a procedure, algorithm, or other function) may vary in sequence or be combined or subdivided. In the following description, for purposes of explanation, numerous specific details are set forth to provide a thorough understanding of various example embodiments. It will be evident to one skilled in the art, however, that the present subject matter may be practiced without these specific details.

[0011] Certain types of Al models (e.g., a machine learning model, which may form all or part of an Al engine) include multiple nodes (e.g., a plurality of nodes, which may be arranged, grouped, or otherwise structured into one or more groups, such as layers). In such Al models, one or more of such nodes may be configured to each produce outputs (e.g., future outputs) based on (e.g., from) inputs (e.g., future inputs) provided to that node. The production of outputs from inputs by a node may be performed using or otherwise based on a generative model of that node. The generative model of a node may be, include, or otherwise be defined by a set of numerical weights, biases, or both (e.g., stored as a tensor), and the generative model of the node may accordingly determine how the node processes inputs received by the node into outputs produced by that same node.

[0012] Regarding types of inputs to a node, for example, a node may be configured to receive directed input (e.g., specifically targeted for processing by the node as part of performing a function of the node, such as an outputting function, which may be shared in common by a group of nodes that include the node), and the directed input may be received by the node from one or more sources (e.g., one or more nodes inside or outside the group of nodes). Additionally, that same node may be configured to receive incidental or observational input (e.g., contextual information, environmental information, or other information not specifically targeted for processing as part of performing the function of the node) from one or more other nodes in its group (e.g., peer nodes, child nodes, or any suitable combination thereof) regarding the outputs of those one or more other nodes in the group.

[0013] According to the systems and methods described herein, a machine (e.g., an Al machine) may facilitate inter-nodal sharing of a generative model by accessing a machine learning model that includes a plurality of nodes. As noted above, each node among the plurality of nodes may be configured to produce future outputs of that node from future inputs to that node based on a generative model of that node. Furthermore, each node among the plurality of nodes may be configured to update its generative model based on past feedback received by that node in response to past outputs of that node produced from past inputs to that node. The machine then associates (e.g., clusters, groups, or otherwise forms an association relationship) a first node among the plurality of nodes with a second node among the plurality of nodes, and the associating (e.g., clustering) of the first and second nodes may be performed based on a comparison of first updates of a first generative model of the first node to second updates of a second generative model of the second node.

[0014] The machine then determines (e.g., selects or generates) a replacement generative model to replace the second generative model of the second node, and this determination is performed in response to the aforementioned associating (e.g., clustering) of the first node with the second node. The machine next replaces (e.g., by substitution) the second generative model of the second node with the determined replacement generative model. This replacing may also be performed in response to the aforementioned associating (e.g., clustering) of the first node with the second node. The replacement generative model configures the second node to produce subsequent outputs of the second node from subsequent inputs to the second node based on the replacement generative model (e.g., instead of continuing to use the replaced second generative model).

[0015] Such sharing of generative models between or among different nodes may provide one or more beneficial effects, including reducing training time for nodes that have less operational experience (e.g., fewer instances of producing outputs from input) than nodes that have more operational experience (e.g., more instances of producing outputs from inputs). For example, in Al models (e.g., machine learning models) able to scale groups of nodes by adding nodes, a newly added node may be quickly updated by replacing its generativemodel with one copied or modified from the generative model of a pre-existing node, thus skipping at least a large portion of the training time and training events experienced by the pre-existing node in evolving its generative model. As another example, in Al models where a group of nodes are all configured to perform the same task or a similar task, all nodes in the group may have their generative models replaced by a single representative (e.g., averaged, including weight-averaged) generative model for improving computational simplicity and efficiency. In addition, storage space, memory allocation, or both, occupied by nodes could be reduced by sharing generative models. For example, a group of one hundred nodes may each contain a pointer to just one of three stored generative models, thus potentially obtaining at least a 90% reduction in storage space, memory footprint, or both, compared to storing all one hundred generative models of all one hundred nodes.

[0016] However, before performing such sharing of a generative model from one node to another node, it may be helpful to ensure that the two nodes have similar evolutionary trajectories in their generative models. Thus, the clustering of two nodes may be based on one or more portions of their sequences of updates to their generative models. Accordingly, clustering the two nodes based on a comparison of one node’s updates (e.g., initial, most recent, or a combination thereof) to the other node’s updates (e.g., initial, most recent, or a combination thereof) may help ensure that the two nodes are not in conflict with each other (e.g., configured to perform divergent or otherwise incompatible tasks) or otherwise out of alignment (e.g., misaligned) with each other, prior to sharing the generative model of one with the other.

[0017] FIG. 1 is a network diagram illustrating a network environment 100 suitable for inter-nodal sharing of one or more generative models, according to some example embodiments. The network environment 100 includes an Al machine 110, a database 115, and devices 130 and 150, all communicatively coupled to each other via a network 190. The Al machine 110, with or without the database 115, may form all or part of a cloud 118 (e.g., a geographically distributed set of multiple machines configured to function as a single server), which may form all or part of a network-based system 105 (e.g., a cloud-based server system configured to provide one or more network-based services to thedevices 130 and 150). The Al machine 110, the database 115, and the devices 130 and 150 may each be implemented in a special-purpose (e.g., specialized) computer system, in whole or in part, as described below with respect to FIG. 12.

[0018] Also shown in FIG. 1 are users 132 and 152. One or both of the users 132 and 152 may be a human user (e.g., a human being), a machine user (e.g., a computer configured by a software program to interact with the device 130 or 150), or any suitable combination thereof (e.g., a human assisted by a machine or a machine supervised by a human). The user 132 is associated with the device 130 and may be a user of the device 130. For example, the device 130 may be a desktop computer, a vehicle computer, a home media system (e.g., a home theater system or other home entertainment system), a tablet computer, a navigational device, a portable media device, a smart phone, or a wearable device (e.g., a smart watch, smart glasses, smart clothing, or smart jewelry) belonging to the user 132. Likewise, the user 152 is associated with the device 150 and may be a user of the device 150. As an example, the device 150 may be a desktop computer, a vehicle computer, a home media system (e.g., a home theater system or other home entertainment system), a tablet computer, a navigational device, a portable media device, a smart phone, or a wearable device (e.g., a smart watch, smart glasses, smart clothing, or smart jewelry) belonging to the user 152.

[0019] Any of the systems or machines (e.g., databases and devices) shown in FIG. 1 may be, include, or otherwise be implemented in a specialpurpose (e.g., specialized or otherwise non-conventional and non-generic) computer that has been modified to perform one or more of the functions described herein for that system or machine (e.g., configured or programmed by special-purpose software, such as one or more software modules of a specialpurpose application, operating system, firmware, middleware, or other software program). For example, a special-purpose computer system able to implement any one or more of the methodologies described herein is discussed below with respect to FIG. 12, and such a special-purpose computer may accordingly be a means for performing any one or more of the methodologies discussed herein. Within the technical field of such special-purpose computers, a special-purposecomputer that has been specially modified (e.g., configured by special-purpose software) by the structures discussed herein to perform the functions discussed herein is technically improved compared to other special-purpose computers that lack the structures discussed herein or are otherwise unable to perform the functions discussed herein. Accordingly, a special-purpose machine configured according to the systems and methods discussed herein provides an improvement to the technology of similar special-purpose machines.

[0020] As used herein, a “database” is a data storage resource and may store data structured in any of various ways, for example, as a text file, a table, a spreadsheet, a relational database (e.g., an object-relational database), a triple store, a hierarchical data store, a document database, a graph database, key -value pairs, or any suitable combination thereof. Moreover, any two or more of the systems or machines illustrated in FIG. 1 may be combined into a single system or machine, and the functions described herein for any single system or machine may be subdivided among multiple systems or machines.

[0021] The network 190 may be any network that enables communication between or among systems, machines, databases, and devices (e.g., between the Al machine 110 and the device 130). Accordingly, the network 190 may be a wired network, a wireless network (e.g., a mobile or cellular network), or any suitable combination thereof. The network 190 may include one or more portions that constitute a private network, a public network (e.g., the Internet), or any suitable combination thereof. Accordingly, the network 190 may include one or more portions that incorporate a local area network (LAN), a wide area network (WAN), the Internet, a mobile telephone network (e.g., a cellular network), a wired telephone network (e.g., a plain old telephone service (POTS) network), a wireless data network (e.g., a WiFi network or WiMax network), or any suitable combination thereof. Any one or more portions of the network 190 may communicate information via a transmission medium. As used herein, “transmission medium” refers to any intangible (e.g., transitory) medium that is capable of communicating (e.g., transmitting) instructions for execution by a machine (e.g., by one or more processors of such a machine), and includes digital or analog communication signals or other intangible media to facilitate communication of such software.

[0022] FIG. 2 is a block diagram illustrating some components of the Al machine, as configured for inter-nodal sharing of one or more generative models, according to some example embodiments. The Al machine 110 is shown as including a machine learning model 210. The machine learning model 210 includes a group 212 of nodes (e.g., each configured to perform a respectively corresponding task, such as to produce future outputs of that node from future inputs to that node based on a generative model of that node, and also to update its generative model based on past feedback received by that node in response to past outputs of that node produced from past inputs to that node). The group 212 of nodes includes multiple nodes (e.g., a plurality of nodes), such as a node 221 (e.g., a first node) and a node 222 (e.g., a second node), which may be a peer of the node 221 (e.g., the first node), a parent of the node 221 (e.g., the first node), a child of the node 221 (e.g., the first node), or a node of some other relationship with the node 221 (e.g., the first node) within the group 212 of nodes.

[0023] As shown in FIG. 2, the machine learning model 210 may form all or part of an app 200 (e.g., a mobile app, a server app, or other computer program) that is stored (e.g., installed) on the Al machine 110 (e.g., responsive to or otherwise as a result of data being received via the network 190, such as from the database 115 or from the device 130). Furthermore, one or more processors 299 (e.g., hardware processors, digital processors, or any suitable combination thereof) may be included (e.g., temporarily or permanently) in the app 200, the machine learning model 210, or any suitable combination thereof.

[0024] Any one or more of the components (e.g., modules) described herein may be implemented using hardware alone (e.g., one or more of the processors 299) or a combination of hardware and software. For example, any component described herein may physically include an arrangement of one or more of the processors 299 (e.g., a subset of or among the processors 299) configured to perform the operations described herein for that component. As another example, any component described herein may include software, hardware, or both, that configure an arrangement of one or more of the processors 299 to perform the operations described herein for that component. Accordingly, different components described herein may include and configuredifferent arrangements of the processors 299 at different points in time or a single arrangement of the processors 299 at different points in time. Each component (e.g., module) described herein is an example of a means for performing the operations described herein for that component. Moreover, any two or more components described herein may be combined into a single component, and the functions described herein for a single component may be subdivided among multiple components. Furthermore, according to various example embodiments, components described herein as being implemented within a single system or machine (e.g., a single device) may be distributed across multiple systems or machines (e.g., multiple devices).

[0025] FIG. 3 is a conceptual diagram illustrating evolution of generative models (e.g., along respective evolutionary paths) in two Al nodes with similar initial sequences of updates to their generative models, according to some example embodiments. As shown in FIG. 3, the node 221 (e.g., the first node) has an initial sequence 321 of updates to its generative model. That is, the generative model of the node 221 (e.g., the first node) has a sequential history of updates, and the initial sequence 321 of updates to the generative model of the node 221 forms an initial portion of an evolutionary path (e.g., a first evolutionary path defined by a first history of sequential updates) taken or otherwise experienced by the generative model of the node 221 (e.g., the first node). The evolutionary path of the generative model of the node 221 (e.g., the first node) is represented in FIG. 3 by a long curved arrow, in which the initial sequence 321 of updates is or includes an initial portion of the total lifetime evolutionary path for the generative model of the node 221 (e.g., the first node).

[0026] As also shown in FIG. 3, the node 222 (e.g., the second node) similarly has an initial sequence 322 of updates to its generative model. That is, the generative model of the node 222 (e.g., the second node) has a sequential history of updates, and the initial sequence 322 of updates to the generative model of the node 222 forms an initial portion of an evolutionary path (e.g., a second evolutionary path defined by a second history of sequential updates) taken or otherwise experienced by the generative model of the node 222 (e.g., the second node). The evolutionary path of the generative model of the node 222 (e.g., the second node) is represented in FIG. 3 by a long curved arrow,where the initial sequence 322 of updates is or includes an initial portion of the total lifetime evolutionary path for the generative model of the node 222 (e.g., the second node).

[0027] In FIG. 3, the broken -line portion of the evolutionary path for the generative model of the node 222 (e.g., the second node) has not yet occurred at the point in time represented (e.g., depicted) in FIG. 3. According to the systems and methods discussed herein, similarity (e.g., including identicality) between the initial sequences 321 and 322 for the nodes 221 and 222 indicate that the generative model of the node 222 (e.g., the second node) will likely evolve like the generative model of the node 221 (e.g., the first node), and further that sharing the generative model of the node 221 (e.g., the first node) with the node 222 (e.g., the second node), whether precisely or approximately, can provide a shortcut to training (e.g., evolving) the generative model of the node 222 (e.g., the second node) and obtain a likely similar evolution of its generative model without the node 222 (e.g., the second node) processing all of the inputs that were processed by the node 221 (e.g., the first node) during its evolution of its generative model. In particular, replacing the generative model of the node 222 (e.g., the second node) with one based on the generative model of the node 221 may obtain the beneficial practical effect of giving the node 222 (e.g., the second node) a large amount of training already experienced by the node 221 (e.g., the first node).

[0028] FIG. 4 is a conceptual diagram illustrating evolution of generative models (e.g., along respective evolutionary paths) in two Al nodes with similar most recent sequences of updates to their generative models, according to some example embodiments. As shown in FIG. 4, the node 221 (e.g., the first node) has a most recent sequence 421 of updates to its generative model. That is, the generative model of the node 221 (e.g., the first node) has its sequential history of updates, and the most recent sequence 421 of updates to the generative model of the node 221 forms a most recent portion of the evolutionary path (e.g., the first evolutionary path) taken or otherwise experienced by the generative model of the node 221 (e.g., the first node). The evolutionary path of the generative model of the node 221 (e.g., the first node) is represented in FIG. 4 by a long curved arrow, in which the most recent sequence 421 of updates is or includes amost recent portion of the total lifetime evolutionary path for the generative model of the node 221 (e.g., the first node).

[0029] As also shown in FIG. 4, the node 222 (e.g., the second node) similarly has a most recent sequence 422 of updates to its generative model. That is, the generative model of the node 222 (e.g., the second node) has its sequential history of updates, and the most recent sequence 422 of updates to the generative model of the node 222 forms a most recent portion of the evolutionary path (e.g., the second evolutionary path) taken or otherwise experienced by the generative model of the node 222 (e.g., the second node). The evolutionary path of the generative model of the node 222 (e.g., the second node) is represented in FIG. 4 by a long curved arrow, where the most recent sequence 422 of updates is or includes a most recent portion of the total lifetime evolutionary path for the generative model of the node 222 (e.g., the second node).

[0030] In FIG. 4, the broken-line portion of the evolutionary path for the generative model of the node 222 (e.g., the second node) did not necessarily occur for the node 222 (e.g., the second node). However, according to the systems and methods discussed herein, similarity (e.g., including identicality) between the most recent sequences 421 and 422 for the nodes 221 and 222 indicate that the generative model of the node 222 (e.g., the second node), going forward, will likely evolve like the generative model of the node 221 (e.g., the first node), and further that sharing the generative model of the node 221 (e.g., the first node) with the node 222 (e.g., the second node), whether precisely or approximately, can provide a shortcut to training (e.g., evolving) the generative model of the node 222 (e.g., the second node) and obtain a likely similar future evolution of its generative model without the node 222 (e.g., the second node) processing all of the inputs that were processed by the node 221 (e.g., the first node) during its evolution of its generative model up to the point in time represented (e.g., depicted) in FIG. 4. In particular, replacing the generative model of the node 222 (e.g., the second node) with one based on the generative model of the node 221 may obtain the beneficial practical effect of giving the node 222 (e.g., the second node) a large amount of training already experienced by the node 221 (e.g., the first node).

[0031] FIGS. 5-11 are flowcharts illustrating operations of the Al machine 110 in performing a method 500 of inter-nodal sharing of a generative model, according to some example embodiments. Operations in the method 500 may be performed by the Al machine 110, using components (e.g., modules) described above with respect to FIG. 2, using one or more processors (e.g., microprocessors or other hardware processors), or using any suitable combination thereof. As shown in FIG. 5, the method 500 includes operations 510, 520, 530, and 540.

[0032] In operation 510, the app 200 accesses the machine learning model 210, which includes the group 212 (e.g., plurality) of nodes. As noted above, each individual node (e.g., nodes 221 and 222) among the group 212 of nodes is configured to produce future outputs of that node from future inputs to that node based on a generative model of that node and also to update the generative model of that node based on past feedback received by that node in response to past outputs of that node produced from past inputs to that node.

[0033] In operation 520, the app 200 associates (e.g., clusters) a first node (e.g., node 221) with a second node (e.g., node 222) among the group 212 of nodes based on a comparison of first updates of a first generative model of the first node (e.g., node 221) to second updates of a second generative model of the second node (e.g., node 222). For example, as noted below regarding various example embodiments, the first updates of the first generative model of the first node (e.g., node 221) may be or include the initial sequence 321 of updates, the most recent sequence 421 of updates, or both, and the second updates of the second generative model of the second node (e.g., node 222) may be or include the initial sequence 322 of updates, the most recent sequence 422 of updates, or both.

[0034] In operation 530, the app 200 determines (e.g., selects or generates) a replacement generative model to replace the second generative model of the second node (e.g., node 222) in response to performance of operation 520 (e.g., the clustering or otherwise associating of the first node with the second node). As noted below regarding various example embodiments, the replacement generative model may be determined based on the first generative model of the first node (e.g., node 221). For example, the replacement generative model forthe second node (e.g., node 222) may be a copy (e.g., an exact copy) of the first generative model of the first node (e.g., node 221). As another example, the replacement generative model for the second node (e.g., node 222) may be a hybrid of the first generative model of the first node (e.g., node 221) and one or more additional generative models of other nodes. In a specific example, the replacement generative model for the second node (e.g., node 222) is a hybrid generative model created (e.g., interpolated or averaged, with or without weighting) based on the first generative model of the first node (e.g., node 221) and the second generative model of the second node (e.g., node 222).

[0035] In operation 540, the app 200 replaces (e.g., overwrites or otherwise substitutes) the second generative model of the second node (e.g., node 222) with the replacement generative model determined in operation 530. As a consequence, the replacement generative model configures the second node (e.g., node 222) to produce its subsequent outputs from its subsequent inputs based on the replacement generative model (e.g., instead of continuing to use the now-replaced second generative model).

[0036] As shown in FIG. 6, in addition to any one or more of the operations previously described, the method 500 may include one or more of operations 630 and 640. Operation 630 may be performed as part (e.g., a precursor task, a subroutine, or a portion) of operation 530, and operation 640 may be performed as part of operation 540.

[0037] In operation 630, as part of the app 200 determining the replacement generative model that will replace the second generative model of the second node (e.g., node 222), the app 200 specifically determines that the first generative model of the first node (e.g., node 221) is the replacement generative model for the second node (e.g., node 222).

[0038] In operation 640, as part of the app 200 replacing the second generative model of the second node (e.g., node 222) with the replacement generative model, the app 200 replaces the second generative model with the first generative model (e.g., with a copy of the first generative model) of the first node (e.g., node 221).

[0039] As shown in FIG. 7, in addition to any one or more of the operations previously described, the method 500 may include one or more of operations 730 and 740. Operation 730 may be performed as part (e.g., a precursor task, a subroutine, or a portion) of operation 530, and operation 740 may be performed as part of operation 540.

[0040] In operation 730, as part of the app 200 determining the replacement generative model that will replace the second generative model of the second node (e.g., node 222), the app 200 generates (e.g., creates) a hybrid generative model based on both the first generative model of the first node (e.g., node 221) and the second generative model of the second node (e.g., node 222), and the app 200 then determines that the generated hybrid generative model is the replacement generative model for the second node (e.g., node 222).

[0041] In some example embodiments, the generating of the hybrid generative model for the second node (e.g., node 222) includes generating an averaged generative model that is generated (e.g., averaged, with or without weighting) based on (e.g., by) averaging together the first and second generative models. In such example embodiments, the determining that the hybrid generative model is the replacement generative model may include determining that the averaged generative model is the replacement generative model.

[0042] In operation 740, as part of the app 200 replacing the second generative model of the second node (e.g., node 222) with the replacement generative model, the app 200 replaces the second generative model with the hybrid generative model generated in operation 730.

[0043] In example embodiments where the generating of the hybrid generative model for the second node (e.g., node 222) includes generating an averaged generative model that is generated (e.g., averaged, with or without weighting) based on (e.g., by) averaging together the first and second generative models, the replacing of the second generative model of the second node (e.g., node 222) may accordingly replace the second generative model with the averaged generative model. According to certain example embodiments, the averaging of the first and second generative models may be weighted (e.g., weighted by how long each node has been in existence, how many inputs eachnode has received, how many outputs each node has produced, or any suitable combination thereof).

[0044] As shown in FIG. 8, in addition to any one or more of the operations previously described, the method 500 may include operation 820. Operation 820 may be performed as part (e.g., a precursor task, a subroutine, or a portion) of operation 520, in which the app 200 associates (e.g., clusters) the first node (e.g., node 221) with the second node (e.g., node 222) based on the comparison of first updates of the first generative model of the first node (e.g., node 221) to second updates of the second generative model of the second node (e.g., node 222).

[0045] In example embodiments that include operation 820, the first updates of the first generative model of the first node include an initial first sequence (e.g., initial sequence 321) of updates within a first history of updates to the first generative model of the first node (e.g., node 221). Similarly, the second updates of the second generative model of the second node include an initial second sequence (e.g., initial sequence 322) of updates within a second history of updates to the second generative model of the second node (e.g., node 222).

[0046] In operation 820, as part of associating (e.g., clustering) the first and second nodes with each other, the app 200 associates (e.g., clusters) the first node (e.g., node 221) with the second node (e.g., node 222) based on a further comparison of the initial first sequence (e.g., initial sequence 321) of updates to the initial second sequence (e.g., initial sequence 322) of updates. In some example embodiments, the further comparison of the initial first sequence of updates to the initial second sequence of updates determines that the initial first sequence of updates and the initial second sequence of updates are similar within a threshold percentage (e.g., a predetermined threshold percentage value) of each other.

[0047] As shown in FIG. 9, in addition to any one or more of the operations previously described, the method 500 may include operation 920. Operation 920 may be performed as part (e.g., a precursor task, a subroutine, or a portion) of operation 520, in which the app 200 associates (e.g., clusters) the first node (e.g., node 221) with the second node (e.g., node 222) based on thecomparison of first updates of the first generative model of the first node (e.g., node 221) to second updates of the second generative model of the second node (e.g., node 222).

[0048] In example embodiments that include operation 920, the first updates of the first generative model of the first node include a most recent first sequence (e.g., most recent sequence 421) of updates within the first history of updates to the first generative model of the first node (e.g., node 221). Similarly, the second updates of the second generative model of the second node include a most recent second sequence (e.g., most recent sequence 422) of updates within the second history of updates to the second generative model of the second node (e.g., node 222).

[0049] In operation 920, as part of associating (e.g., clustering) the first and second nodes with each other, the app 200 associates (e.g., clusters) the first node (e.g., node 221) with the second node (e.g., node 222) based on a further comparison of the most recent first sequence (e.g., most recent sequence 421) of updates to the most recent second sequence (e.g., most recent sequence 422) of updates. In some example embodiments, the further comparison of the most recent first sequence of updates to the most recent second sequence of updates determines that the most recent first sequence of updates and the most recent second sequence of updates are similar within a threshold percentage (e.g., a predetermined threshold percentage value) of each other.

[0050] As shown in FIG. 10, in addition to any one or more of the operations previously described, the method 500 may include operation 1020. Operation 1020 may be performed as part (e.g., a precursor task, a subroutine, or a portion) of operation 520, in which the app 200 associates (e.g., clusters) the first node (e.g., node 221) with the second node (e.g., node 222) based on the comparison of first updates of the first generative model of the first node (e.g., node 221) to second updates of the second generative model of the second node (e.g., node 222).

[0051] In example embodiments that include operation 1020, similarity between the current generative models for the first and second nodes is a basis (e.g., an additional basis or the sole basis) for associating the first and secondnodes for inter-nodal sharing of a generative model (e.g., the first generative model of the first node).

[0052] In operation 1020, as part of associating (e.g., clustering) the first and second nodes with each other, the app 200 associates (e.g., clusters) the first node (e.g., node 221) with the second node (e.g., node 221) based on a further comparison of the first generative model of the first node (e.g., node 221) to the second generative model of the second node (e.g., node 222). In some example embodiments, the further comparison of the first generative model of the first node (e.g., node 221) to the second generative model of the second node (e.g., node 222) determines that the first generative model and the second generative model are within a threshold percentage (e.g., a predetermined threshold percentage value) of each other.

[0053] As shown in FIG. 11, in addition to any one or more of the operations previously described, the method 500 may include one or more of operations 1130 and 1140. Operation 1130 may be performed as part (e.g., a precursor task, a subroutine, or a portion) of operation 530 or after operation 530, and operation 1140 may be performed as part of operation 540 or after operation 540.

[0054] In operation 1130, the app 200 determines (e.g., selects or chooses) that the replacement generative model is to replace the first generative model of the first node (e.g., node 221) in response to the clustering of the first node (e.g., node 221) with the second node (e.g., node 222) in operation 520. Accordingly, in addition to the second node (e.g., node 222) having its generative model replaced with the replacement generative model, the first node (e.g., node 221) also will have its generative model replaced with the replacement generative model (e.g., a hybrid generative model).

[0055] In operation 1140, the app 200 replaces (e.g., overwrites or otherwise substitutes) the first generative model of the first node (e.g., node 221) with the replacement generative model determined in operation 530. As a consequence, the replacement generative model configures the first node (e.g., node 221) to produce its subsequent outputs from its subsequent inputs based on the replacement generative model (e.g., instead of continuing to use the now- replaced first generative model).

[0056] According to various example embodiments, one or more of the methodologies described herein may facilitate sharing of one or more generative models between or among different Al nodes. Moreover, one or more of the methodologies described herein may facilitate reductions in training time for one or more Al nodes within a machine learning model or other Al engine. Hence, one or more of the methodologies described herein may facilitate scaling of machine learning models and computationally efficient training of new or younger Al nodes, as well as computational simplicity and efficiency from sharing hybrid generative models between or among multiple Al nodes, compared to capabilities of pre-existing systems and methods.

[0057] When these effects are considered in aggregate, one or more of the methodologies described herein may obviate a need for certain efforts or resources that otherwise would be involved in inter-nodal sharing of generative models, scaling machine learning models or other Al engines, training Al nodes, improving computational efficiencies in machine learning models or other Al engines, or any suitable combination thereof. Efforts expended by a user in obtaining such benefits may be reduced by use of (e.g., reliance upon) a specialpurpose machine that implements one or more of the methodologies described herein. Computing resources used by one or more systems or machines (e.g., within the network environment 100) may similarly be reduced (e.g., compared to systems or machines that lack the structures discussed herein or are otherwise unable to perform the functions discussed herein). Examples of such computing resources include processor cycles, network traffic, computational capacity, main memory usage, graphics rendering capacity, graphics memory usage, data storage capacity, power consumption, and cooling capacity.

[0058] FIG. 12 is a block diagram illustrating components of a machine 1200, according to some example embodiments, able to read instructions 1224 from a machine-readable medium 1222 (e.g., a non-transitory machine-readable medium, a machine-readable storage medium, a computer-readable storage medium, or any suitable combination thereof) and perform any one or more of the methodologies discussed herein, in whole or in part. Specifically, FIG. 12 shows the machine 1200 in the example form of a computer system (e.g., a computer) within which the instructions 1224 (e.g., software, a program, anapplication, an applet, an app, or other executable code) for causing the machine 1200 to perform any one or more of the methodologies discussed herein may be executed, in whole or in part.

[0059] In alternative embodiments, the machine 1200 operates as a standalone device or may be communicatively coupled (e.g., networked) to other machines. In a networked deployment, the machine 1200 may operate in the capacity of a server machine or a client machine in a server-client network environment, or as a peer machine in a distributed (e.g., peer-to-peer) network environment. The machine 1200 may be a server computer, a client computer, a personal computer (PC), a tablet computer, a laptop computer, a netbook, a cellular telephone, a smart phone, a set-top box (STB), a personal digital assistant (PDA), a web appliance, a network router, a network switch, a network bridge, or any machine capable of executing the instructions 1224, sequentially or otherwise, that specify actions to be taken by that machine. Further, while only a single machine is illustrated, the term “machine” shall also be taken to include any collection of machines that individually or jointly execute the instructions 1224 to perform all or part of any one or more of the methodologies discussed herein.

[0060] The machine 1200 includes a processor 1202 (e.g., one or more central processing units (CPUs), one or more graphics processing units (GPUs), one or more digital signal processors (DSPs), one or more application specific integrated circuits (ASICs), one or more radio-frequency integrated circuits (RFICs), one or more tensor processing units (TPUs), one or more neural processing units (NPUs), one or more vision processing units (VPUs), one or more machine-learning accelerators, one or more artificial-intelligence accelerators, one or more neuromorphic processors, one or more quantum processors, or any suitable combination thereof), a main memory 1204, and a static memory 1206, which are configured to communicate with each other via a bus 1208. The processor 1202 contains solid-state digital microcircuits (e.g., electronic, optical, biological, or any suitable combination thereof) that are configurable, temporarily or permanently, by some or all of the instructions 1224 such that the processor 1202 is configurable to perform any one or more of the methodologies described herein, in whole or in part. For example, a set of oneor more microcircuits of the processor 1202 may be configurable to execute one or more modules (e.g., software modules) described herein.

[0061] In some example embodiments, the processor 1202 is a multicore CPU (e.g., a dual-core CPU, a quad-core CPU, an 8-core CPU, or a 128-core CPU) within which each of multiple cores behaves as a separate processor that is able to perform any one or more of the methodologies discussed herein, in whole or in part. Parallel processing on one or more multicore processors may be implemented via one or more suitable architectures, such as superscalar, very long instruction word (VLIW), vector processing, or single-instruction / multiple- data (SIMD), which allow each core to run separate instruction streams concurrently. A processor may be emulated in software, running on a physical processor, as a virtual processor or virtual circuit. The virtual processor may behave like an independent processor but is implemented in software rather than hardware. Although the beneficial effects described herein may be provided by the machine 1200 with at least the processor 1202, these same beneficial effects may be provided by a different kind of machine that contains no processors (e.g., a purely mechanical system, a purely hydraulic system, or a hybrid mechanical- hydraulic system), if such a processor-less machine is configured to perform one or more of the methodologies described herein.

[0062] The machine 1200 may further include a graphics display 1210 (e.g., a plasma display panel (PDP), a light emitting diode (LED) display, a liquid crystal display (LCD), a projector, a cathode ray tube (CRT), or any other display capable of displaying graphics or video). The machine 1200 may also include an alphanumeric input device 1212 (e.g., a keyboard or keypad), a pointer input device 1214 (e.g., a mouse, a touchpad, a touchscreen, a trackball, a joystick, a stylus, a motion sensor, an eye tracking device, a data glove, or other pointing instrument), a data storage 1216, an audio generation device 1218 (e.g., a sound card, an amplifier, a speaker, a headphone jack, or any suitable combination thereof), and a network interface device 1220.

[0063] The data storage 1216 (e.g., a data storage device) includes the machine-readable medium 1222 (e.g., a tangible and non-transitory machine- readable storage medium) on which are stored the instructions 1224 embodying any one or more of the methodologies or functions described herein. Theinstructions 1224 may also reside, completely or at least partially, within the main memory 1204, within the static memory 1206, within the processor 1202 (e.g., within the processor’s cache memory), or any suitable combination thereof, before or during execution thereof by the machine 1200. Accordingly, the main memory 1204, the static memory 1206, and the processor 1202 may be considered machine-readable media (e.g., tangible and non-transitory machine- readable media). The instructions 1224 may be transmitted or received over the network 190 via the network interface device 1220. For example, the network interface device 1220 may communicate the instructions 1224 using any one or more transfer protocols (e.g., hypertext transfer protocol (HTTP)).

[0064] In some example embodiments, the machine 1200 may be a portable computing device (e.g., a smart phone, a tablet computer, or a wearable device) and may have one or more additional input components 1230 (e.g., sensors or gauges). Examples of such input components 1230 include an image input component (e.g., one or more cameras), an audio input component (e.g., one or more microphones), a direction input component (e.g., a compass), a location input component (e.g., a global positioning system (GPS) receiver), an orientation component (e.g., a gyroscope), a motion detection component (e.g., one or more accelerometers), an altitude detection component (e.g., an altimeter), a temperature input component (e.g., a thermometer), and a gas detection component (e.g., a gas sensor). Input data gathered by any one or more of these input components 1230 may be accessible and available for use by any of the modules described herein (e.g., with suitable privacy notifications and protections, such as opt-in consent or opt-out consent, implemented in accordance with user preference, applicable regulations, or any suitable combination thereof).

[0065] As used herein, the term “memory” refers to a machine-readable medium able to store data temporarily or permanently and may be taken to include, but not be limited to, random-access memory (RAM), read-only memory (ROM), buffer memory, flash memory, and cache memory. While the machine-readable medium 1222 is shown in an example embodiment to be a single medium, the term “machine-readable medium” should be taken to include a single medium or multiple media (e.g., a centralized or distributed database, orassociated caches and servers) able to store instructions. The term “machine- readable medium” shall also be taken to include any medium, or combination of multiple media, that is capable of carrying (e.g., storing or communicating) the instructions 1224 for execution by the machine 1200, such that the instructions 1224, when executed by one or more processors of the machine 1200 (e.g., processor 1202), cause the machine 1200 to perform any one or more of the methodologies described herein, in whole or in part. Accordingly, a “machine- readable medium” refers to a single storage apparatus or device, as well as cloud-based storage systems or storage networks that include multiple storage apparatus or devices. The term “machine-readable medium” shall accordingly be taken to include, but not be limited to, one or more tangible and non- transitory data repositories (e.g., data volumes) in the example form of a solid- state memory chip, an optical disc, a magnetic disc, or any suitable combination thereof.

[0066] A “non-transitory” machine-readable medium, as used herein, specifically excludes propagating signals per se. According to various example embodiments, the instructions 1224 for execution by the machine 1200 can be communicated via a carrier medium (e.g., a machine-readable carrier medium). Examples of such a carrier medium include a non-transient carrier medium (e.g., a non-transitory machine-readable storage medium, such as a solid-state memory that is physically movable from one place to another place) and a transient carrier medium (e.g., a carrier wave or other propagating signal that communicates the instructions 1224).

[0067] Certain example embodiments are described herein as including modules. Modules may constitute software modules (e.g., code stored or otherwise embodied in a machine-readable medium or in a transmission medium), hardware modules, or any suitable combination thereof. A “hardware module” is a tangible (e.g., non-transitory) physical component (e.g., a set of one or more processors) capable of performing certain operations and may be configured or arranged in a certain physical manner. In various example embodiments, one or more computer systems or one or more hardware modules thereof may be configured by software (e.g., an application or portion thereof) asa hardware module that operates to perform operations described herein for that module.

[0068] In some example embodiments, a hardware module may be implemented mechanically, electronically, hydraulically, or any suitable combination thereof. For example, a hardware module may include dedicated circuitry or logic that is permanently configured to perform certain operations. A hardware module may be or include a special-purpose processor, such as a field programmable gate array (FPGA) or an ASIC. A hardware module may also include programmable logic or circuitry that is temporarily configured by software to perform certain operations. As an example, a hardware module may include software encompassed within a CPU or other programmable processor. It will be appreciated that the decision to implement a hardware module mechanically, hydraulically, in dedicated and permanently configured circuitry, or in temporarily configured circuitry (e.g., configured by software) may be driven by cost and time considerations.

[0069] Accordingly, the phrase “hardware module” should be understood to encompass a tangible entity that may be physically constructed, permanently configured (e.g., hardwired), or temporarily configured (e.g., programmed) to operate in a certain manner or to perform certain operations described herein. Furthermore, as used herein, the phrase “hardware-implemented module” refers to a hardware module. Considering example embodiments in which hardware modules are temporarily configured (e.g., programmed), each of the hardware modules need not be configured or instantiated at any one instance in time. For example, where a hardware module includes a CPU configured by software to become a special-purpose processor, the CPU may be configured as respectively different special-purpose processors (e.g., each included in a different hardware module) at different times. Software (e.g., a software module) may accordingly configure one or more processors, for example, to become or otherwise constitute a particular hardware module at one instance of time and to become or otherwise constitute a different hardware module at a different instance of time.

[0070] Hardware modules can provide information to, and receive information from, other hardware modules. Accordingly, the described hardware modules may be regarded as being communicatively coupled. Wheremultiple hardware modules exist contemporaneously, communications may be achieved through signal transmission (e.g., over circuits and buses) between or among two or more of the hardware modules. In embodiments in which multiple hardware modules are configured or instantiated at different times, communications between such hardware modules may be achieved, for example, through the storage and retrieval of information in memory structures to which the multiple hardware modules have access. For example, one hardware module may perform an operation and store the output of that operation in a memory (e.g., a memory device) to which it is communicatively coupled. A further hardware module may then, at a later time, access the memory to retrieve and process the stored output. Hardware modules may also initiate communications with input or output devices, and can operate on a resource (e.g., a collection of information from a computing resource).

[0071] The various operations of example methods described herein may be performed, at least partially, by one or more processors that are temporarily configured (e.g., by software) or permanently configured to perform the relevant operations. Whether temporarily or permanently configured, such processors may constitute processor-implemented modules that operate to perform one or more operations or functions described herein. As used herein, “processor- implemented module” refers to a hardware module in which the hardware includes one or more processors. Accordingly, the operations described herein may be at least partially processor-implemented, hardware-implemented, or both, since a processor is an example of hardware, and at least some operations within any one or more of the methods discussed herein may be performed by one or more processor-implemented modules, hardware-implemented modules, or any suitable combination thereof.

[0072] Moreover, such one or more processors may perform operations in a “cloud computing” environment or as a service (e.g., within a “software as a service” (SaaS) implementation). For example, at least some operations within any one or more of the methods discussed herein may be performed by a group of computers (e.g., as examples of machines that include processors), with these operations being accessible via a network (e.g., the Internet) and via one or more appropriate interfaces (e.g., an application program interface (API)). Theperformance of certain operations may be distributed among the one or more processors, whether residing only within a single machine or deployed across a number of machines. In some example embodiments, the one or more processors or hardware modules (e.g., processor-implemented modules) may be located in a single geographic location (e.g., within a home environment, an office environment, or a server farm). In other example embodiments, the one or more processors or hardware modules may be distributed across a number of geographic locations.

[0073] Throughout this specification, plural instances may implement components, operations, or structures described as a single instance. Although individual operations of one or more methods are illustrated and described as separate operations, one or more of the individual operations may be performed concurrently, and nothing requires that the operations be performed in the order illustrated. Structures and their functionality presented as separate components and functions in example configurations may be implemented as a combined structure or component with combined functions. Similarly, structures and functionality presented as a single component may be implemented as separate components and functions. These and other variations, modifications, additions, and improvements fall within the scope of the subject matter herein.

[0074] Some portions of the subject matter discussed herein may be presented in terms of algorithms or symbolic representations of operations on data stored as bits or binary digital signals within a memory (e.g., a computer memory or other machine memory). Such algorithms or symbolic representations are examples of techniques used by those of ordinary skill in the data processing arts to convey the substance of their work to others skilled in the art. As used herein, an “algorithm” is a self-consi stent sequence of operations or similar processing leading to a desired result. In this context, algorithms and operations involve physical manipulation of physical quantities. Typically, but not necessarily, such quantities may take the form of electrical, magnetic, or optical signals capable of being stored, accessed, transferred, combined, compared, or otherwise manipulated by a machine. It is convenient at times, principally for reasons of common usage, to refer to such signals using words such as “data,” “content,” “bits,” “values,” “elements,” “symbols,” “characters,”“terms,” “numbers,” “numerals,” or the like. These words, however, are merely convenient labels and are to be associated with appropriate physical quantities.

[0075] Unless specifically stated otherwise, discussions herein using words such as “accessing,” “processing,” “detecting,” “computing,” “calculating,” “determining,” “generating,” “presenting,” “displaying,” or the like refer to actions or processes performable by a machine (e.g., a computer) that manipulates or transforms data represented as physical (e.g., electronic, magnetic, or optical) quantities within one or more memories (e.g., volatile memory, non-volatile memory, or any suitable combination thereof), registers, or other machine components that receive, store, transmit, or display information. Furthermore, unless specifically stated otherwise, the terms “a” or “an” are herein used, as is common in patent documents, to include one or more than one instance. Finally, as used herein, the conjunction “or” refers to a nonexclusive “or,” unless specifically stated otherwise.

[0076] The following enumerated descriptions describe various examples of methods, machine-readable media, and systems (e.g., machines, devices, or other apparatus) discussed herein. Any one or more features of an example, taken in isolation or combination, should be considered as being within the disclosure of this application.

[0077] A first example provides a method comprising: accessing, by one or more processors, a machine learning model that includes a plurality of nodes, each node among the plurality of nodes being configured to produce future outputs of that node from future inputs to that node based on a generative model of that node and to update the generative model of that node based on past feedback received by that node in response to past outputs of that node produced from past inputs to that node; clustering, by the one or more processors, a first node with a second node among the plurality of nodes based on a comparison of first updates of a first generative model of the first node to second updates of a second generative model of the second node;determining, by the one or more processors, a replacement generative model to replace the second generative model of the second node in response to the clustering of the first node with the second node; and replacing, by the one or more processors, the second generative model of the second node with the determined replacement generative model, the replacement generative model configuring the second node to produce subsequent outputs of the second node from subsequent inputs to the second node based on the replacement generative model.

[0078] A second example provides a method according to the first example, wherein: the determining of the replacement generative model for the second node determines that the first generative model of the first node is the replacement generative model; and the replacing of the second generative model of the second node replaces the second generative model with the first generative model.

[0079] A third example provides a method according to the first example or the second example, wherein: the determining of the replacement generative model for the second node includes generating a hybrid generative model based on the first and second generative models, and determining that the hybrid generative model is the replacement generative model; and the replacing of the second generative model of the second node replaces the second generative model with the hybrid generative model.

[0080] A fourth example provides a method according to the third example, wherein: the generating of the hybrid generative model for the second node includes generating an averaged generative model based on the first and second generative models; the determining that the hybrid generative model is the replacement generative model determines that the averaged generative model is the replacement generative model; andthe replacing of the second generative model of the second node replaces the second generative model with the averaged generative model.

[0081] A fifth example provides a method according to any of the first through fourth examples, wherein: the first updates of the first generative model of the first node include an initial first sequence of updates within a first history of updates to the first generative model; the second updates of the second generative model of the second node include an initial second sequence of updates within a second history of updates to the second generative model; and the clustering of the first node with the second node is based on a further comparison of the initial first sequence of updates to the initial second sequence of updates.

[0082] A sixth example provides a method according to the fifth example, wherein: the further comparison of the initial first sequence of updates to the initial second sequence of updates determines that the initial first sequence of updates and the initial second sequence of updates are similar within a threshold percentage of each other.

[0083] A seventh example provides a method according to any of the first through sixth examples, wherein: the first updates of the first generative model of the first node include a most recent first sequence of updates within a first history of updates to the first generative model; the second updates of the second generative model of the second node include a most recent second sequence of updates within a second history of updates to the second generative model; and the clustering of the first node with the second node is based on a further comparison of the most recent first sequence of updates to the most recent second sequence of updates.

[0084] An eighth example provides a method according to the seventh example, wherein: the further comparison of the most recent first sequence of updates to the most recent second sequence of updates determines that the most recent first sequence of updates and the most recent second sequence of updates are similar within a threshold percentage of each other.

[0085] A ninth example provides a method according to any of the first through eighth examples, wherein: the clustering of the first node with the second node is based on a further comparison of the first generative model of the first node to the second generative model of the second node.

[0086] A tenth example provides a method according to the ninth example, wherein: the further comparison of the first generative model of the first node to the second generative model of the second node determines that the first generative model and the second generative model are within a threshold percentage of each other.

[0087] An eleventh example provides a method according to any of the first through tenth examples, further comprising: determining that the replacement generative model is to replace the first generative model of the first node in response to the clustering of the first node with the second node; and replacing the first generative model of the first node with the replacement generative model, the replacement generative model configuring the first node to produce subsequent outputs from subsequent inputs based on the replacement generative model.

[0088] A twelfth example provides a machine-readable medium (e.g., a non-transitory machine-readable storage medium) comprising instructions that, when executed by one or more processors of a machine, cause the machine to perform operations comprising:accessing a machine learning model that includes a plurality of nodes, each node among the plurality of nodes being configured to produce future outputs of that node from future inputs to that node based on a generative model of that node and to update the generative model of that node based on past feedback received by that node in response to past outputs of that node produced from past inputs to that node; clustering a first node with a second node among the plurality of nodes based on a comparison of first updates of a first generative model of the first node to second updates of a second generative model of the second node; determining a replacement generative model to replace the second generative model of the second node in response to the clustering of the first node with the second node; and replacing the second generative model of the second node with the determined replacement generative model, the replacement generative model configuring the second node to produce subsequent outputs of the second node from subsequent inputs to the second node based on the replacement generative model.

[0089] A thirteenth example provides a machine-readable medium according to the twelfth example, wherein: the determining of the replacement generative model for the second node determines that the first generative model of the first node is the replacement generative model; and the replacing of the second generative model of the second node replaces the second generative model with the first generative model.

[0090] A fourteenth example provides a machine-readable medium according to the twelfth example or the thirteenth example, wherein: the determining of the replacement generative model for the second node includes generating a hybrid generative model based on the first and second generative models, and determining that the hybrid generative model is the replacement generative model; and the replacing of the second generative model of the second node replaces the second generative model with the hybrid generative model.

[0091] A fifteenth example provides a machine-readable medium according to any of the twelfth through fourteenth examples, wherein: the first updates of the first generative model of the first node include an initial first sequence of updates within a first history of updates to the first generative model; the second updates of the second generative model of the second node include an initial second sequence of updates within a second history of updates to the second generative model; and the clustering of the first node with the second node is based on a further comparison of the initial first sequence of updates to the initial second sequence of updates.

[0092] A sixteenth example provides a machine-readable medium according to any of the twelfth through fifteenth examples, wherein: the first updates of the first generative model of the first node include a most recent first sequence of updates within a first history of updates to the first generative model; the second updates of the second generative model of the second node include a most recent second sequence of updates within a second history of updates to the second generative model; and the clustering of the first node with the second node is based on a further comparison of the most recent first sequence of updates to the most recent second sequence of updates.

[0093] A seventeenth example provides a system (e.g., a computer system that includes one or more computers, devices, or other machines) comprising: one or more processors; and a memory storing instructions that, when executed by at least one processor among the one or more processors, cause the system to perform operations comprising: accessing a machine learning model that includes a plurality of nodes, each node among the plurality of nodes being configured to produce future outputs of that node from future inputs to that node based on a generative model of that nodeand to update the generative model of that node based on past feedback received by that node in response to past outputs of that node produced from past inputs to that node; clustering a first node with a second node among the plurality of nodes based on a comparison of first updates of a first generative model of the first node to second updates of a second generative model of the second node; determining a replacement generative model to replace the second generative model of the second node in response to the clustering of the first node with the second node; and replacing the second generative model of the second node with the determined replacement generative model, the replacement generative model configuring the second node to produce subsequent outputs of the second node from subsequent inputs to the second node based on the replacement generative model.

[0094] An eighteenth example provides a system according to the seventeenth example, wherein: the determining of the replacement generative model for the second node determines that the first generative model of the first node is the replacement generative model; and the replacing of the second generative model of the second node replaces the second generative model with the first generative model.

[0095] A nineteenth example provides a system according to the seventeenth example or the eighteenth example, wherein: the determining of the replacement generative model for the second node includes generating a hybrid generative model based on the first and second generative models, and determining that the hybrid generative model is the replacement generative model; and the replacing of the second generative model of the second node replaces the second generative model with the hybrid generative model.

[0096] A twentieth example provides a system according to any of the seventeenth through nineteenth examples, wherein:the first updates of the first generative model of the first node include an initial first sequence of updates within a first history of updates to the first generative model; the second updates of the second generative model of the second node include an initial second sequence of updates within a second history of updates to the second generative model; and the clustering of the first node with the second node is based on a further comparison of the initial first sequence of updates to the initial second sequence of updates. A twenty -first example provides a carrier medium carrying machine-readable instructions for controlling a machine to carry out the operations (e.g., method operations) performed in any one of the previously described examples.

Claims

CLAIMSWhat is claimed is:

1. A method comprising: accessing, by one or more processors, a machine learning model that includes a plurality of nodes, each node among the plurality of nodes being configured to produce future outputs of that node from future inputs to that node based on a generative model of that node and to update the generative model of that node based on past feedback received by that node in response to past outputs of that node produced from past inputs to that node; clustering, by the one or more processors, a first node with a second node among the plurality of nodes based on a comparison of first updates of a first generative model of the first node to second updates of a second generative model of the second node; determining, by the one or more processors, a replacement generative model to replace the second generative model of the second node in response to the clustering of the first node with the second node; and replacing, by the one or more processors, the second generative model of the second node with the determined replacement generative model, the replacement generative model configuring the second node to produce subsequent outputs of the second node from subsequent inputs to the second node based on the replacement generative model.

2. The method of claim 1, wherein: the determining of the replacement generative model for the second node determines that the first generative model of the first node is the replacement generative model; and the replacing of the second generative model of the second node replaces the second generative model with the first generative model.

3. The method of claim 1, wherein: the determining of the replacement generative model for the second node includes generating a hybrid generative model based on the first and second generative models, and determining that the hybrid generative model is the replacement generative model; and the replacing of the second generative model of the second node replaces the second generative model with the hybrid generative model.

4. The method of claim 3, wherein: the generating of the hybrid generative model for the second node includes generating an averaged generative model based on the first and second generative models; the determining that the hybrid generative model is the replacement generative model determines that the averaged generative model is the replacement generative model; and the replacing of the second generative model of the second node replaces the second generative model with the averaged generative model.

5. The method of claim 1, wherein: the first updates of the first generative model of the first node include an initial first sequence of updates within a first history of updates to the first generative model; the second updates of the second generative model of the second node include an initial second sequence of updates within a second history of updates to the second generative model; and the clustering of the first node with the second node is based on a further comparison of the initial first sequence of updates to the initial second sequence of updates.

6. The method of claim 5, wherein: the further comparison of the initial first sequence of updates to the initial second sequence of updates determines that the initial firstsequence of updates and the initial second sequence of updates are similar within a threshold percentage of each other.

7. The method of claim 1, wherein: the first updates of the first generative model of the first node include a most recent first sequence of updates within a first history of updates to the first generative model; the second updates of the second generative model of the second node include a most recent second sequence of updates within a second history of updates to the second generative model; and the clustering of the first node with the second node is based on a further comparison of the most recent first sequence of updates to the most recent second sequence of updates.

8. The method of claim 7, wherein: the further comparison of the most recent first sequence of updates to the most recent second sequence of updates determines that the most recent first sequence of updates and the most recent second sequence of updates are similar within a threshold percentage of each other.

9. The method of claim 1, wherein: the clustering of the first node with the second node is based on a further comparison of the first generative model of the first node to the second generative model of the second node.

10. The method of claim 9, wherein: the further comparison of the first generative model of the first node to the second generative model of the second node determines that the first generative model and the second generative model are within a threshold percentage of each other.

11. The method of claim 1, further comprising: determining that the replacement generative model is to replace the first generative model of the first node in response to the clustering of the first node with the second node; and replacing the first generative model of the first node with the replacement generative model, the replacement generative model configuring the first node to produce subsequent outputs from subsequent inputs based on the replacement generative model.

12. A non-transitory machine-readable storage medium comprising instructions that, when executed by one or more processors of a machine, cause the machine to perform operations comprising: accessing a machine learning model that includes a plurality of nodes, each node among the plurality of nodes being configured to produce future outputs of that node from future inputs to that node based on a generative model of that node and to update the generative model of that node based on past feedback received by that node in response to past outputs of that node produced from past inputs to that node; clustering a first node with a second node among the plurality of nodes based on a comparison of first updates of a first generative model of the first node to second updates of a second generative model of the second node; determining a replacement generative model to replace the second generative model of the second node in response to the clustering of the first node with the second node; and replacing the second generative model of the second node with the determined replacement generative model, the replacement generative model configuring the second node to produce subsequent outputs of the second node from subsequent inputs to the second node based on the replacement generative model.

13. The non-transitory machine-readable storage medium of claim 12, wherein: the determining of the replacement generative model for the second node determines that the first generative model of the first node is the replacement generative model; and the replacing of the second generative model of the second node replaces the second generative model with the first generative model.

14. The non-transitory machine-readable storage medium of claim 12, wherein: the determining of the replacement generative model for the second node includes generating a hybrid generative model based on the first and second generative models, and determining that the hybrid generative model is the replacement generative model; and the replacing of the second generative model of the second node replaces the second generative model with the hybrid generative model.

15. The non-transitory machine-readable storage medium of claim 12, wherein: the first updates of the first generative model of the first node include an initial first sequence of updates within a first history of updates to the first generative model; the second updates of the second generative model of the second node include an initial second sequence of updates within a second history of updates to the second generative model; and the clustering of the first node with the second node is based on a further comparison of the initial first sequence of updates to the initial second sequence of updates.

16. The non-transitory machine-readable storage medium of claim 12, wherein: the first updates of the first generative model of the first node include a most recent first sequence of updates within a first history of updates to the first generative model; the second updates of the second generative model of the second node include a most recent second sequence of updates within a second history of updates to the second generative model; andthe clustering of the first node with the second node is based on a further comparison of the most recent first sequence of updates to the most recent second sequence of updates.

17. A system comprising: one or more processors; and a memory storing instructions that, when executed by at least one processor among the one or more processors, cause the system to perform operations comprising: accessing a machine learning model that includes a plurality of nodes, each node among the plurality of nodes being configured to produce future outputs of that node from future inputs to that node based on a generative model of that node and to update the generative model of that node based on past feedback received by that node in response to past outputs of that node produced from past inputs to that node; clustering a first node with a second node among the plurality of nodes based on a comparison of first updates of a first generative model of the first node to second updates of a second generative model of the second node; determining a replacement generative model to replace the second generative model of the second node in response to the clustering of the first node with the second node; and replacing the second generative model of the second node with the determined replacement generative model, the replacement generative model configuring the second node to produce subsequent outputs of the second node from subsequent inputs to the second node based on the replacement generative model.

18. The system of claim 17, wherein: the determining of the replacement generative model for the second node determines that the first generative model of the first node is the replacement generative model; andthe replacing of the second generative model of the second node replaces the second generative model with the first generative model.

19. The system of claim 17, wherein: the determining of the replacement generative model for the second node includes generating a hybrid generative model based on the first and second generative models, and determining that the hybrid generative model is the replacement generative model; and the replacing of the second generative model of the second node replaces the second generative model with the hybrid generative model.

20. The system of claim 17, wherein: the first updates of the first generative model of the first node include an initial first sequence of updates within a first history of updates to the first generative model; the second updates of the second generative model of the second node include an initial second sequence of updates within a second history of updates to the second generative model; and the clustering of the first node with the second node is based on a further comparison of the initial first sequence of updates to the initial second sequence of updates.

Images