Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
DETAILED ACTION
The Action is responsive to the preliminary amendments filed on 6/10/2025 and the Application filed on 5/6/2025. Claims 1-20 are pending claims. Claims 1, 10, and 19 are written in independent form.
Priority
Acknowledgment is made of a claim for priority as a continuation of PCT/CN2022/130311, filed 11/07/2022 and designating the US.
Claim Objections
Claims 1, 10, 19, and 20 are objected to because of the following informalities:
Claims 1, 10, and 19 appear to recite a typographical error by referring to “the training data” when “preset training data” was previously recited. The intended language is understood as reciting “the preset training data” since no other training data was mentioned in the claims.
Claim 20 appears to recite a typographical error by reciting “wherein the training…further cause an apparatus to:” when “an apparatus” was previously recited in Claim 19 upon which Claim 20 depends. The intended language is understood as reciting “wherein the training…further causes the apparatus to:”.
Appropriate correction is required.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claim(s) 1, 7, 10, 16, and 19 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Krishnamoorthy et al. (U.S. Pre-Grant Publication No. 2021/0042623, hereinafter referred to as Krishnamoorthy).
Regarding Claim 1:
Krishnamoorthy teaches a method comprising:
Loading, in response to a first version number input by a user, a first embedding representation corresponding to the first version number from a storage medium into a memory;
Krishnamoorthy teaches a “user input to check-out a first image processing model” is received and causes the system to “extract the model…from the MVS repository(s) 260 to the local user repository 230 or user device 104” for modifying the respective model (Para. [0144] & Fig. 8) and “The system may identify operational discrete elements of a remote interaction type comprising commands such as “publish” for publishing a model (e.g., a mutated model) to the system 210, “search” for searching/selecting models in the MVS repository 260, “pull” for downloading model files, etc.” (Para. [0102]).Krishnamoorthy further teaches using version numbers by teaching “embodiments of the invention also provide an electronic system for versioning machine-learning neural-network based image processing models and identifying and tracking mutations in hyper parameters amongst versions of image processing models.” (Para. [0001]) and “the operational discrete element 522a indicating the action to be performed is “construct,” the identifier discrete element 522c indicating the location/name of the new model/program comprises “m2”, the operator discrete element 522b indicating the preposition type relational context is “from”, a keyword discrete element 522d indicating the existing model is “m1”,” (Para. [0096]).
Training the first embedding representation based on preset training data to obtain a second embedding representation;
Krishnamoorthy teaches “the system may receive a second user input associated with a request to store a second image processing model at least one hosted model versioning system repository. Here, the system may then receive user input queries comprising a plurality of discrete input language elements having the operational discrete element of “add”, such as “add model2 to imagenew”.” (Para. [0145]) where “The system may then initiate training of the model by providing training images to the model. Here, the system typically constructs metadata for the model and stores it at image processing DL model metadata 232, and constructs parameterized storage via the PAS system 280 (e.g., highly compressed storage of the model having altered weigh parameter objects and hyperparameters during training, as described by FIG. 8) and stores it at parameter archival storage 240, and constructs model artifacts and stores it in a predetermined folder structure in the Git Repository 234.” (Para. [0061]) thereby teaching after modifying the model through training using locally stored images, a second model is obtained having different weights and/or hyperparameters that is then stored at the hosted model versioning system (MVS) Repository(s) 260.
Determining a second version number of the second embedding representation based on the first version number and a scenario associated with the training data; and
Krishnamoorthy teaches “the operational discrete element 522a indicating the action to be performed is “construct,” the identifier discrete element 522c indicating the location/name of the new model/program comprises “m2”, the operator discrete element 522b indicating the preposition type relational context is “from”, a keyword discrete element 522d indicating the existing model is “m1”,” (Para. [0096]).
Storing the second embedding representation and the second version number on the storage medium.
Krishnamoorthy teaches “In response to processing these user inputs via the DL model query command application 214, the system invokes an “add” function (e.g., stored in the computer readable instructions 154) and transmits the models files (e.g., of the image processing model 301) to the hosted MVS repositories 260, via the server 250.” (Para. [0069]) and “The system may then storing the second image processing model at the at least one hosted model versioning system repository by storing only (i) the weigh parameter objects, and (ii) the second hyper parameter.” (Para. [0149]).
Regarding Claim 7:
Kirshnamoorthy further teaches:
wherein the storing the second embedding representation and the second version number on the storage medium comprises:
storing the second embedding representation on the storage medium in a multi-level differential storage manner.
Krishnamoorthy teaches “embodiments of the invention also provide an electronic system for versioning machine-learning neural-network based image processing models and…a parameter archival storage system” (Para. [0001]).
Regarding Claim 10:
Some of the limitations herein are similar to some or all of the limitations as recited in Claim 1.
Krishnamoorthy further teaches:
An apparatus (Krishnamoorthy - Para. [0153]) comprising:
A memory configured to store instructions; and (Krishnamoorthy - Para. [0154])
One or more processors coupled to the memory and configured to execute the instructions to cause the apparatus to perform steps. (Krishnamoorthy - Para. [0153])
Regarding Claim 16:
All of the limitations herein are similar to some or all of the limitations as recited in Claim 7.
Regarding Claim 19:
Some of the limitations herein are similar to some or all of the limitations as recited in Claim 1.
Krishnamoorthy further teaches:
A computer program product comprising computer-executable instructions that are stored on a non-transitory computer-readable storage medium that, when executed by a processor, cause an apparatus to perform steps. (Krishnamoorthy - Paras. [0153]-[0154])
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 2-3, 8, 9, 11-12, 17, 18, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Krishnamoorthy and further in view of Huszar (U.S. Patent No. 11,410,076).
Regarding Claim 2:
Krishnamoorthy explicitly teaches all of the elements of the claimed invention as recited above except:
Wherein the training the first embedding representation to obtain a second embedding representation comprises:
In a training process of the first embedding representation, obtaining a first intermediate version of the first embedding representation based on a preset first time interval, wherein the first time interval is on a daily basis;
when a latest obtained first intermediate version meets a preset first evaluation condition, storing the latest obtained first intermediate version on the storage medium; and
when a preset training end condition is met, ending training to obtain the second embedding representation.
However, in the related field of endeavor of storing data in a version control datastore, Huszar teaches:
Wherein the training the first embedding representation to obtain a second embedding representation comprises:
In a training process of the first embedding representation, obtaining a first intermediate version of the first embedding representation based on a preset first time interval, wherein the first time interval is on a daily basis;
Huszar teaches “the loss value for each task in the datastore can be calculated on a regular basis (e.g., as a scheduled event, hourly, daily, monthly, and/or the like) and stored (e.g., in version control datastore 415). In this way modified (e.g., newly trained) parameters may replace existing parameters if the loss is improved (e.g., the new loss is less than a previous loss value).” (Col. 12 Line 59 – Col. 13 Line 7).
when a latest obtained first intermediate version meets a preset first evaluation condition, storing the latest obtained first intermediate version on the storage medium; and
Huszar teaches “the loss value for each task in the datastore can be calculated on a regular basis (e.g., as a scheduled event, hourly, daily, monthly, and/or the like) and stored (e.g., in version control datastore 415). In this way modified (e.g., newly trained) parameters may replace existing parameters if the loss is improved (e.g., the new loss is less than a previous loss value).” (Col. 12 Line 59 – Col. 13 Line 7).
when a preset training end condition is met, ending training to obtain the second embedding representation.
Huszar teaches “the loss value for each task in the datastore can be calculated on a regular basis (e.g., as a scheduled event, hourly, daily, monthly, and/or the like) and stored (e.g., in version control datastore 415). In this way modified (e.g., newly trained) parameters may replace existing parameters if the loss is improved (e.g., the new loss is less than a previous loss value).” (Col. 12 Line 59 – Col. 13 Line 7).
Thus, it would have been obvious to one of ordinary skill in the art, having the teachings of Huszar and Krishnamoorthy at the time that the claimed invention was effectively filed, to have modified the systems, methods, and computer program products for an electronic system for management of image processing model database, as taught by Krishnamoorthy, with the sharing of parameters allowed for transfer learning between models, as taught by Huszar.
One would have been motivated to make such combination because Huszar teaches “The shared set of parameters can allow for transfer learning between models and corresponding tasks and/or datasets. Thus allowing each model to benefit from multiple views on the data available to the organization” (Col. 1 Lines 49-67).
Regarding Claim 3:
Huszar and Krishnamoorthy further teach:
wherein the training the first embedding representation to obtain a second embedding representation comprises:
in the training process of the first embedding representation, obtaining a second intermediate version of the first embedding representation based on a preset second time interval, wherein the second time interval is on an hourly basis or a minute basis;
Huszar teaches “the loss value for each task in the datastore can be calculated on a regular basis (e.g., as a scheduled event, hourly, daily, monthly, and/or the like) and stored (e.g., in version control datastore 415). In this way modified (e.g., newly trained) parameters may replace existing parameters if the loss is improved (e.g., the new loss is less than a previous loss value).” (Col. 12 Line 59 – Col. 13 Line 7).
determining a first difference between a latest obtained second intermediate version and a previous second intermediate version; and
Krishnamoorthy teaches determining differences between versions by teaching “the second image processing model can be stored in a compressed parameterized manner by merely storing the (i) only the weigh parameter objects (floating point weights) of the second image processing model that differ from that of the first image processing model, and (ii) first mutant neural network layer component that is linked/mapped to the first image processing model,” (Para. [0115]).
when the first difference meets a preset second evaluation condition, backing up the latest obtained second intermediate version to the memory.
Krishnamoorthy teaches “the system may process/analyze a first plurality of weights associated with the plurality of first convolutional neural network layers of the first image processing model and a corresponding second plurality of weights associated with the plurality of second convolution neural network layers associated with the second image processing model. The system may then determine altered weights in the second plurality of weights that deviate from the corresponding first plurality of weights. The system may then map the altered weights in the second plurality of weights with the corresponding first plurality of weights and the corresponding plurality of first convolutional neural network layers. Indeed, here the constructed weigh parameter objects for the second image processing model are the only altered weights” (Para. [0115]).
Regarding Claim 8:
Huszar and Krishnamoorthy further teach:
wherein the storing the second embedding representation on the storage medium in a multi-level differential storage manner comprises:
determining a fifth embedding representation from the second embedding representation, wherein the fifth embedding representation is an embedding representation that is in the second embedding representation and that is changed relative to the first embedding representation;
Krishnamoorthy teaches “the system is configured to store the second image processing model by: based on mapping the mutations in hyper parameters between the first plurality of hyper parameters of the first image processing model and the second plurality of hyper parameters associated with the second image processing model, discarding the hierarchical linked architecture of the second image processing model; and storing the second image processing model at the at least one hosted model versioning system repository by storing only (i) the weigh parameter objects, and (ii) mapped mutations in hyper parameters.” (Para. [0010]) and “Each image processing model may further comprise one or more learning rate components. The learning rate is a hyperparameter that controls how much to change the model in response to the estimated error each time the model weights are updated. It determines to what extent newly acquired information overrides old information, i.e., indicates learning rate decay or momentum.” (Para. [0125]) where “The system may then storing the second image processing model at the at least one hosted model versioning system repository by storing only (i) the weigh parameter objects, and (ii) the second hyper parameter.” (Para. [0149]). Therefore Krishnamoorthy teaches determining multiple representations of changes relative to the first model and storing the representations.
storing the fifth embedding representation on the storage medium; and
Krishnamoorthy teaches “the system is configured to store the second image processing model by: based on mapping the mutations in hyper parameters between the first plurality of hyper parameters of the first image processing model and the second plurality of hyper parameters associated with the second image processing model, discarding the hierarchical linked architecture of the second image processing model; and storing the second image processing model at the at least one hosted model versioning system repository by storing only (i) the weigh parameter objects, and (ii) mapped mutations in hyper parameters.” (Para. [0010]) and “Each image processing model may further comprise one or more learning rate components. The learning rate is a hyperparameter that controls how much to change the model in response to the estimated error each time the model weights are updated. It determines to what extent newly acquired information overrides old information, i.e., indicates learning rate decay or momentum.” (Para. [0125]) where “The system may then storing the second image processing model at the at least one hosted model versioning system repository by storing only (i) the weigh parameter objects, and (ii) the second hyper parameter.” (Para. [0149]). Therefore Krishnamoorthy teaches determining multiple representations of changes relative to the first model and storing the representations.
establishing, based on a first address and a second address, a storage mapping table corresponding to the second embedding representation,
Huszar teaches “the parameter datastore can includes at least one look-up table that stores a plurality of parameters as real-valued vectors having a dimensionality corresponding to a number of input parameters associated with at least one of the plurality of tasks, and the version control datastore can include at least one look-up table that stores data associated with the plurality of parameters as real-valued vectors having a dimensionality corresponding to a number of input parameters associated with at least one of the plurality of tasks.” (Col. 3 Lines 49-58) and “generating a new parameter or set of parameters (e.g., by adding a new task and/or by changing a machine learning model associated with an existing task) to be stored in a parameter datastore (e.g., the parameter datastore 110)” (Col. 8 Lines 4-25).
wherein the first address is an address, on the storage medium, of an embedding representation that is in the second embedding representation and that is not changed relative to the first embedding representation, and
Krishnamoorthy teaches “the second image processing model can be stored in a compressed parameterized manner by merely storing the (i) only the weigh parameter objects (floating point weights) of the second image processing model that differ from that of the first image processing model, and (ii) first mutant neural network layer component that is linked/mapped to the first image processing model, instead of the entire hierarchical linked architecture framework of the second image processing model.” (Para. [0115]).
the second address is an address, on the storage medium, of the fifth embedding representation.
Krishnamoorthy teaches “the second image processing model can be stored in a compressed parameterized manner by merely storing the (i) only the weigh parameter objects (floating point weights) of the second image processing model that differ from that of the first image processing model, and (ii) first mutant neural network layer component that is linked/mapped to the first image processing model, instead of the entire hierarchical linked architecture framework of the second image processing model.” (Para. [0115]).
Regarding Claim 9:
Huszar and Krishnamoorthy further teach:
in response to a user request for comparing the third embedding representation and a fourth embedding representation, separately performing dimensionality reduction on the third embedding representation and the fourth embedding representation, to obtain a first dimensionality reduction vector and a second dimensionality reduction vector;
Huszar teaches “The new parameter vector (θ) can be a vector having a dimensionality corresponding to a number of input parameters associated with the task (e.g., task T.sub.i) and/or to a default dimensionality as defined by the version control datastore (e.g., version control datastore 415) and/or the parameter datastore (e.g., the parameter datastore 110). The values associated with the new parameter vector (θ) can be set to a default value (e.g., 0, 1, and the like) and/or defined by the task. In addition, the dimensionality of the new parameter vector (θ) can be increased beyond that corresponding to a number of input parameters associated with the task and/or default dimensionality by zero-padding the new parameter vector (θ).” (Col. 8 Lines 4-25) and
determining a second difference between the third embedding representation and the fourth embedding representation; and
Krishnamoorthy teaches determining differences between versions by teaching “the second image processing model can be stored in a compressed parameterized manner by merely storing the (i) only the weigh parameter objects (floating point weights) of the second image processing model that differ from that of the first image processing model, and (ii) first mutant neural network layer component that is linked/mapped to the first image processing model,” (Para. [0115]).
displaying the first dimensionality reduction vector, the second dimensionality reduction vector, and the second difference.
Krishnamoorthy teaches “In some embodiments, the user 102 may be one or more individuals or entities that may either provide static UI images (e.g., via the image capture device 180), e.g., for model training, request selection and check-out of image processing models, input queries for search and selection of models, initiate mutation of models, view displayed models, etc.” (Para. [0036]) and “an electronic system for versioning machine-learning neural-network based image processing models and identifying and tracking mutations in hyper parameters amongst versions of image processing models” (Para. [0050]).
Regarding Claim 11:
All of the limitations herein are similar to some or all of the limitations as recited in Claim 2.
Regarding Claim 12:
All of the limitations herein are similar to some or all of the limitations as recited in Claim 3.
Regarding Claim 17:
All of the limitations herein are similar to some or all of the limitations as recited in Claim 8.
Regarding Claim 18:
All of the limitations herein are similar to some or all of the limitations as recited in Claim 9.
Regarding Claim 20:
All of the limitations herein are similar to some or all of the limitations as recited in Claim 2.
Claim(s) 4 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Haszar and Krishnamoorthy and further in view of Agarwal et al. (U.S. Pre-Grant Publication No. 2021/0405984, hereinafter referred to as Agarwal).
Regarding Claim 4:
Haszar and Krishnamoorthy explicitly teach all of the elements of the claimed invention as recited above except:
wherein the training the first embedding representation to obtain a second embedding representation comprises:
when one of the latest obtained first intermediate version does not meet the first evaluation condition, or the first difference does not meet the second evaluation condition, selecting one of a rollback version from the second intermediate version in the memory or the first intermediate version on the storage medium according to a preset version rollback rule; and
continuing training based on the rollback version.
However, in the related field of endeavor of versioned machine-learning models, Agarwal teaches:
wherein the training the first embedding representation to obtain a second embedding representation comprises:
when one of the latest obtained first intermediate version does not meet the first evaluation condition, or the first difference does not meet the second evaluation condition, selecting one of a rollback version from the second intermediate version in the memory or the first intermediate version on the storage medium according to a preset version rollback rule; and
Agarwal teaches “deployed version 116 fails to satisfy a threshold level.” and “ in response to the determining, computer system 101 rolls back deployed version 116 of machine-learning model 115 to a different version (e.g., version 115a or 115b) stored in database 110.” (Para. [0024]).
continuing training based on the rollback version.
Agarwal teaches “ in response to the determining, computer system 101 rolls back deployed version 116 of machine-learning model 115 to a different version (e.g., version 115a or 115b) stored in database 110.” (Para. [0024]) thereby teaching using the rollback version for all future training.
Thus, it would have been obvious to one of ordinary skill in the art, having the teachings of Agarwal, Huszar, and Krishnamoorthy at the time that the claimed invention was effectively filed, to have modified the systems, methods, and computer program products for an electronic system for management of image processing model database, as taught by Krishnamoorthy, and the sharing of parameters allowed for transfer learning between models, as taught by Huszar, with the roll back of model versions to alternative versions, as taught by Agarwal.
One would have been motivated to make such combination because Agarwal teaches “the disclosed techniques may provide mitigation for poor performance of a new version of a model by providing a means to roll the new version back to an alternate version that may avoid or address the performance issue.” (Para. [0025]).
Regarding Claim 13:
All of the limitations herein are similar to some or all of the limitations as recited in Claim 4.
Claim(s) 5-6 and 14-15 are rejected under 35 U.S.C. 103 as being unpatentable over Haszar and Krishnamoorthy and further in view of Badawy et al. (U.S. Pre-Grant Publication No. 2022/0269990, hereinafter referred to as Badawy).
Regarding Claim 5:
Haszar and Krishnamoorthy explicitly teach all of the elements of the claimed invention as recited above except:
loading a first nearest neighbor graph corresponding to the first embedding representation from the storage medium into the memory; and
the training the first embedding representation to obtain a second embedding represetation comprises:
in the training process of the first embedding representation, dynamically updating the first nearest neighbor graph based on the second time interval, to obtain a nearest neighbor graph corresponding to each second intermediate version; and
after the second embedding representation is obtained, determining a second nearest neighbor graph corresponding to the second embedding representation.
However, in the related field of endeavor of tracking changes to data over time, Badawy teaches:
loading a first nearest neighbor graph corresponding to the first embedding representation from the storage medium into the memory; and
Badawy teaches “If an incremental training event has occurred (Y branch of STEP 440), a second dataset may be obtained (STEP 450). This second dataset may comprise data obtained from the enterprise or may be data obtained from the enterprise (or determined from data obtained from the enterprise) subsequent to a time at which the initial dataset (e.g., the first dataset) used to train the machine learning model was obtained or determined, or comprise other data.” (Para. [0094]).
the training the first embedding representation to obtain a second embedding representation comprises:
in the training process of the first embedding representation, dynamically updating the first nearest neighbor graph based on the second time interval, to obtain a nearest neighbor graph corresponding to each second intermediate version; and
Badawy teaches “this machine learning model 572 may be trained or tested based on data produced or otherwise associated with the enterprise environment 500. Accordingly, at some time interval, (e.g., when identity management data 554 is obtained from identity management system 550 or identity management graph 565 is updated based on such identity management data 554), identity management system 560 may determine an associated dataset of machine learning training data 594. Machine learning model trainer 574 can then utilize an enterprise dataset 554 or machine learning training dataset 594 to train machine learning model 572.” (Para. [0123]).
after the second embedding representation is obtained, determining a second nearest neighbor graph corresponding to the second embedding representation.
Badawy teaches “This results in a machine learning model that is updated with the latest data patterns of the newer data and does not suffer from performance loss. The incremental training of machine learning models can be achieved by several techniques. In weight modification the weights of the model used to make a decision are modified. In a tree-based ensemble model, appending or a regeneration approach works in which the models can be appended with new nodes or trees trained on new data. In a replacement technique, randomly or statically selected nodes or trees from the machine learning model may be replaced with new nodes or trees trained on new data.” (Para. [0023])
Thus, it would have been obvious to one of ordinary skill in the art, having the teachings of Badawy, Huszar, and Krishnamoorthy at the time that the claimed invention was effectively filed, to have modified the systems, methods, and computer program products for an electronic system for management of image processing model database, as taught by Krishnamoorthy, and the sharing of parameters allowed for transfer learning between models, as taught by Huszar, with the updating of models without having to retrain the entire model from scratch , as taught by Badawy.
One would have been motivated to make such combination because Badawy teaches “Embodiments may result in an updated, more accurate machine learning model without having to retrain the entire model from scratch. Another advantage is that based on the amount of new data or the change of patterns in the new data, it can be determined how much importance needs to be given to the new instances so that a resulting model has a preference between old patterns and new patterns.” (Para. [0036]).
Regarding Claim 6:
Badawy, Haszar, and Krishnamoorthy, further teach:
wherein the determining a first difference between a latest obtained second intermediate version and a previous second intermediate version comprises:
obtaining a third nearest neighbor graph and a fourth nearest neighbor graph, wherein the third nearest neighbor graph is a nearest neighbor graph corresponding to the latest obtained second intermediate version, and the fourth nearest neighbor graph is a nearest neighbor graph corresponding to the previous second intermediate version;
Badawy teaches “by utilizing these graph embeddings, changes in various specific aspects and drifts in the identity graphs may be detected. For example, certain nodes or edges of the graph may be associated with identities, entitlements or roles. Certain relationships or edges of the graph may be associated with connection weights between the nodes representing theses identity management artifacts (e.g., identities, roles, entitlements, etc.). By scoping the graph to certain nodes or edges and embedding only these scoped nodes or edges of the identity graphs, the application of a drift detection model to the embeddings of a first identity graph and a second identity graph may be able to detect drift in particular identity management artifacts (e.g., identities, roles, entitlements, etc.) or relationships between those identity management artifacts.” (Para. [0018])
determining a changed node in the third nearest neighbor graph based on the fourth nearest neighbor graph;
Badawy teaches “by utilizing these graph embeddings, changes in various specific aspects and drifts in the identity graphs may be detected. For example, certain nodes or edges of the graph may be associated with identities, entitlements or roles. Certain relationships or edges of the graph may be associated with connection weights between the nodes representing theses identity management artifacts (e.g., identities, roles, entitlements, etc.). By scoping the graph to certain nodes or edges and embedding only these scoped nodes or edges of the identity graphs, the application of a drift detection model to the embeddings of a first identity graph and a second identity graph may be able to detect drift in particular identity management artifacts (e.g., identities, roles, entitlements, etc.) or relationships between those identity management artifacts.” (Para. [0018])
determining neighbor change information and node change information of each changed node,
wherein the neighbor change information comprises at least one of
a neighbor change quantity,
Badawy teaches “ drift detection model 174 may be based on Kolmogorov-Smirnov windowing (KSWIN)” where “The Kolmogorov-Smirnov statistic quantifies a distance between the empirical distribution function of the sample and the cumulative distribution function of the reference distribution, or between the empirical distribution functions of two samples.” (Para. [0075])
a neighbor change ratio, or
Badawy teaches “the similarity weight could be based on a count of the similarity (e.g., overlap or intersection of identities) between the two entitlements divided by the union of identities. For instance, the similarity could be defined as the ratio between a number of identities having both entitlements joined by the edge to the number of identities that have either one (e.g., including both) of the two entitlements.” (Para. [0117]).
a local neighbor similarity score, and
Badawy teaches “the similarity weight could be based on a count of the similarity (e.g., overlap or intersection of identities) between the two entitlements divided by the union of identities. For instance, the similarity could be defined as the ratio between a number of identities having both entitlements joined by the edge to the number of identities that have either one (e.g., including both) of the two entitlements.” (Para. [0117]).
the node change information comprises at least one of
a node offset direction and a node offset distance; and
Badawy teaches “a drift detection model may be applied to the first and second dataset to determine the drift measure. In one embodiment, the drift detection model may be trained or otherwise determined based on the first dataset (e.g., the dataset that was used to train the machine learning model at the first point in time). This training may, for example, including the determination of one or more metrics associated with the first dataset that may be used in the determination of drift relative to a second dataset. In this manner, the drift detection model can be tailored specifically to the associated machine learning model (or models) trained on that same dataset (or a portion thereof). Examples of such drift detection models include drift detection models based on a Probably Approximately Correct (PAC) learning model, Adaptive Windowing, Hoeffding's bounds, Kolmogorov-Smirnov windowing, Wasserstein distance, Kullback-Leibler divergence, Jenson-Shannon method, T-test, box plots, histograms, or other types of drift detection models.” (Para. [0015]).
determining the first difference between the latest obtained second intermediate version and the previous second intermediate version based on the neighbor change information and the node change information of each changed node.
Badawy teaches “the drift detection model 174 may be based on Wasserstein distance. This is a distance metric defined between two probability distributions in same metric space M. Intuitively, if each distribution is viewed as a unit amount of earth (soil) piled on M, the metric is the minimum “cost” of turning one pile into the other, which is assumed to be the amount of earth that needs to be moved times the mean distance it has to be moved. Because of this analogy, the metric is known in computer science as the earth mover's distance.” (Para. [0077])
Regarding Claim 14:
All of the limitations herein are similar to some or all of the limitations as recited in Claim 5.
Regarding Claim 15:
All of the limitations herein are similar to some or all of the limitations as recited in Claim 6.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Thirumalai (U.S. Pre-Grant Publication No. 2020/0311037) teaches state information at a file location named with a version number. In an example, a data store stores replica state information having a file location named with a first version number. A second version number is received from a consensus protocol, and the file location of the state information is rename with the second version number. The replica state information is updated at the file location named with the second version number while servicing requests for client data.
Martin et al. (U.S. Pre-Grant Publication No. 2020/0034365) teaches updating a target table with changes introduced into a source table. New data records are stored into partitions of the source table. An incremental update module copies the new data records from the source table into the target table, thereby assigning a new target partition ID contained within a first value domain to each copied data record. A view specifies a set of visible partition IDs as a union of the first value domain and of a current target partition ID set and allows the execution of database queries selectively on target table partitions having assigned a target partition ID that is element of the visible partition ID set. A batch update module performs an atomic batch update operation that comprises copying partitions of the source table into a respective new partition in the target table.
Chopra et al. (U.S. Pre-Grant Publication No. 2021/0334169) teaches vendor-neutral models of vendors' application resources are described. A host outputs capabilities of data protection operations which are specified by a vendor of an application that is installed on the host. The host inputs a vendor-neutral version of a data protection operation, based on any of the capabilities, for a resource of the application. The host uses a vendor-neutral model of the resource of the application to perform the vendor-neutral version of the data protection operation on the application resource.The reference further teaches “The data object's management system can use a transaction log backup file to re-apply the changes made by committed transitions that are not materialized in a data object and roll back the changes to a data object that were made by uncommitted transactions.” (Para. [0003]).
Upadhyay et al. (U.S. Pre-Grant Publication No. 2022/0121523) teaches extracting backup metadata, comprising committed change numbers, system change numbers, types, database version identifiers, and details, for backup copies. System generates data structure, comprising start system change numbers, based on minimums of system committed change numbers, end system change numbers and maximums of system change numbers, database version identifiers, and identifiers including details, for the backup copies. System identifies at least first type of backup copies as candidate dependent backup copies and at least second type of backup copies as candidate preceding backup copies. System outputs identifiers of candidate dependent backup copy, corresponding dependency, and candidate preceding backup copy, based on identifying same database version identifier of candidate dependent backup copy and candidate preceding backup copy, and maximum of end system change numbers of candidate preceding backup copies which is less than or equal to start system change number and/or end system change number of candidate dependent backup copy.
Foreign Publication CN 111130842B teaches a construction method of a dynamic network map database reflecting network multi-dimensional resources. The method collects and stores real-time information of each node and link in a network view, including node CPU usage, node memory usage, node name, Node IP, link delay, link bandwidth, link packet loss rate, use in-memory database to build distributed graph database, add multi-level storage mechanism and incremental index design, perform data aging and store historical data to disk. The invention divides the nodes in the network topology into three levels: a centralized management center, a distributed proxy measurement point and a common end node. The centralized management center issues measurement instructions to the proxy measurement point, and the proxy measurement point reports the measurement data to the management center. Realize the update of dynamic network graph database.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ROBERT F MAY whose telephone number is (571)272-3195. The examiner can normally be reached Monday-Friday 9:30am to 6pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Boris Gorney can be reached on 571-270-5626. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/ROBERT F MAY/Examiner, Art Unit 2154 5/30/2026
/BORIS GORNEY/Supervisory Patent Examiner, Art Unit 2154
Prosecution Insights