1
DETAILED ACTION
1. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
2. This communication is in response to the Applicant’s submission filed 08 February 2024, where:
Claims 1-15 are pending.
Claims 1-15 are rejected.
Foreign priority is claimed to FI 20235204, filed 17 February 2023. A certified copy of this paper has been filed 11 March 2024. Accordingly, receipt is acknowledged of certified copies of papers required by 37 CFR 1.55.
Information Disclosure Statement
3. Information disclosure statements were submitted on 10 May 2024 and 09 August 2024. The submissions comply with the provisions of 37 CFR 1.97. Accordingly, the Examiner considered the information disclosure statements.
Specification
4. The title of the invention is not descriptive. A new title is required that is clearly indicative of the invention to which the claims are directed.
Claim Rejections - 35 U.S.C. § 101
5. 35 U.S.C. § 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
6. Claims 1-15 are rejected under 35 U.S.C. § 101 because the claimed invention is directed to an abstract idea without significantly more.
Claim 1 recites an apparatus, which is a product, and thus one of the statutory categories of patentable subject matter. (35 U.S.C. § 101).
However, under Step 2A Prong One, the claim recites the limitations of “validating the machine learning model based on the data structure and the current condition to obtain a validation result,” and “providing the validation result in response to the request to validate the machine learning model.” These activities of “validating” and “providing” contain limitations that can practically be performed in the human mind, including, for example, observations, evaluations, judgments, and opinions, and accordingly, are a mental process, (MPEP § 2106.04(a)(2) sub III), which is one of the groupings of abstract ideas. (MPEP § 2106.04(a)(2)).
The claim also recites more details or specifics of the abstract idea of “providing,” “wherein the data structure of the machine learning model comprises at least one of metadata of the machine learning model or context data of the machine learning model,” and accordingly, is merely more specific to the abstract idea. Thus, the claim recites an abstract idea.
Under Step 2A Prong Two, the claim as a whole is not integrated into a practical application, because the additional elements recited in the claim beyond the identified judicial exception include “at least one processor; and at least one memory storing instructions that, when executed by the at least one processor,” which is the use of generic computer components (at least one processor, at least one memory) to implement the abstract idea, (MPEP § 2106.05(f)), that does not serve to integrate the abstract idea into a practical application. The claim also recites a “machine learning model,” which is recited at a high-level of generality, and accordingly, is a generic computer component used to implement the abstract idea, (MPEP § 2106.05(f)), that does not serve to integrate the abstract idea into a practical application.
The claim also recites the limitation of “receiving a request to validate a machine learning model,” which is a pre-processing, insignificant extra-solution activity of collecting data, (MPEP § 2106.05(g)), that does not serve to integrate the abstract idea into a practical application. The claim further recites more details or specifics to the additional element of “receiving,” “wherein the request comprises a data structure of the machine learning model and an indication of a current condition under which inference by the machine learning model is to be executed,” and accordingly, is merely more specific to the additional element. Therefore, the claim is directed to the abstract idea.
Finally, under Step 2B, the additional elements, taken alone or in combination, do not represent significantly more than the abstract idea itself. The additional elements recited in the claim beyond the identified judicial exception include “at least one processor; and at least one memory storing instructions that, when executed by the at least one processor,” which is the use of generic computer components (at least one processor, at least one memory) to implement the abstract idea, (MPEP § 2106.05(f)), that does not amount to significantly more than the abstract idea. The claim also recites a “machine learning model,” which is recited at a high-level of generality, and accordingly, is a generic computer component used to implement the abstract idea, (MPEP § 2106.05(f)), that does not amount to significantly more than the abstract idea.
The claim also recites the limitation of “receiving a request to validate a machine learning model,” which is a well-understood, routine, and conventional activity of receiving data over a network, (MPEP § 2106.05(d) sub II.i), that does not amount to significantly more than the abstract idea. The claim further recites more details or specifics to the additional element of “receiving,” “wherein the request comprises a data structure of the machine learning model and an indication of a current condition under which inference by the machine learning model is to be executed,” and accordingly, is merely more specific to the additional element. Therefore, claim 1 is subject-matter ineligible.
Claim 11 recites an method, which is a process, and thus one of the statutory categories of patentable subject matter. (35 U.S.C. § 101).
However, under Step 2A Prong One, the claim recites the limitations of “validating the machine learning model based on the data structure and the current condition to obtain a validation result,” and “providing the validation result in response to the request to validate the machine learning model.” These activities of “validating” and “providing” contain limitations that can practically be performed in the human mind, including, for example, observations, evaluations, judgments, and opinions, and accordingly, are a mental process, (MPEP § 2106.04(a)(2) sub III), which is one of the groupings of abstract ideas. (MPEP § 2106.04(a)(2)).
The claim also recites more details or specifics of the abstract idea of “providing,” “wherein the data structure of the machine learning model comprises at least one of metadata of the machine learning model or context data of the machine learning model,” and accordingly, is merely more specific to the abstract idea. Thus, the claim recites an abstract idea.
Under Step 2A Prong Two, the claim as a whole is not integrated into a practical application, because the additional elements recited in the claim beyond the identified judicial exception include a “machine learning model,” which is recited at a high-level of generality, and accordingly, is a generic computer component used to implement the abstract idea, (MPEP § 2106.05(f)), that does not serve to integrate the abstract idea into a practical application.
The claim also recites the limitation of “receiving a request to validate a machine learning model,” which is a pre-processing, insignificant extra-solution activity of collecting data, (MPEP § 2106.05(g)), that does not serve to integrate the abstract idea into a practical application. The claim further recites more details or specifics to the additional element of “receiving,” “wherein the request comprises a data structure of the machine learning model and an indication of a current condition under which inference by the machine learning model is to be executed,” and accordingly, is merely more specific to the additional element. Therefore, the claim is directed to the abstract idea.
Finally, under Step 2B, the additional elements, taken alone or in combination, do not represent significantly more than the abstract idea itself. The additional elements recited in the claim beyond the identified judicial exception include a “machine learning model,” which is recited at a high-level of generality, and accordingly, is a generic computer component used to implement the abstract idea, (MPEP § 2106.05(f)), that does not amount to significantly more than the abstract idea.
The claim also recites the limitation of “receiving a request to validate a machine learning model,” which is a well-understood, routine, and conventional activity of receiving data over a network, (MPEP § 2106.05(d) sub II.i), that does not amount to significantly more than the abstract idea. The claim further recites more details or specifics to the additional element of “receiving,” “wherein the request comprises a data structure of the machine learning model and an indication of a current condition under which inference by the machine learning model is to be executed,” and accordingly, is merely more specific to the additional element. Therefore, claim 11 is subject-matter ineligible.
Claim 2 depends directly or indirectly from claim 1. Claim 12 depends directly or indirectly from claim 11. The claims recite more details or specifics to the abstract idea of “validating,” where “the validating without using the machine learning model,” which is merely more specific to the abstract idea. Also, the additional elements of the claim does not serve to integrate the abstract idea into integrated into a practical application, (see MPEP § 2106.04(d)), nor do the additional elements amount to significantly more than the abstract idea, (MPEP § 2106.05 sub I; see also MPEP § 2106.05(a) – (h)), and thus, the claim recites no more than the abstract idea. Therefore, claims 2 and 12 are subject-matter ineligible.
Claim 3 depends directly or indirectly from claim 1. Claim 13 depends directly or indirectly from claim 11. The claims further recite “providing at least one of an update of the data structure or an update of the machine learning model in response to the request to validate the machine learning model,” which is a post-processing, insignificant extra-solution activity of transmitting feedback to the validation result, (MPEP § 2106.05(g)), that does not serve to integrate the abstract idea into a practical application. Also, the limitation is a well-understood, routine, and conventional activity of transmitting data over a network, (MPEP § 2106.05(d) sub II.i), that does not amount to significantly more than the abstract idea. Therefore, claims 3 and 13 are subject-matter ineligible.
Claims 4 and 5 depend directly or indirectly from claim 1. Claims 14 and 15 depend directly or indirectly from claim 11. The claims recite more details or specifics to the additional element of “receiving a request,” (claims 4 and 14: wherein the data structure of the machine learning model comprises an identifier of a machine learning model, and at least one of the following: static information on the machine learning model; dynamic information on the machine learning model; or secure information on the machine learning model; . . .”; claims 5 and 15: wherein at least one of the following: the condition under which the machine learning model was trained comprises at least one of the following: a network at which the machine learning model was trained; a radio parameter under which the machine learning model was trained; a radio condition under which the machine learning model was trained; or a parameter of a terminal under which the machine learning model was trained; or the structure of the training data used to train the machine learning model comprises at least one of the following: an input parameter of the machine learning model; a range of values of the input parameter in the training data; or a distribution of the values of the input parameter in the training data; or the secure information is encrypted in the data structure), and accordingly, are merely more specific to the additional element. Therefore, claims 4, 5, 14, and 15 are subject-matter ineligible.
Claim 6 recites an apparatus, which is a product, and thus one of the statutory categories of patentable subject matter. (35 U.S.C. § 101).
However, under Step 2A Prong One, the claim recites the limitation of “deciding whether or not to perform inference by the machine learning model based on the validation result,.” The activity of “deciding” contains a limitation that can practically be performed in the human mind, including, for example, observations, evaluations, judgments, and opinions, and accordingly, is a mental process, (MPEP § 2106.04(a)(2) sub III), which is one of the groupings of abstract ideas. (MPEP § 2106.04(a)(2)).
The claim also recites more details or specifics of the abstract idea of “deciding,” “wherein the data structure of the machine learning model comprises at least one of metadata of the machine learning model or context data of the machine learning model,” and accordingly, is merely more specific to the abstract idea. Thus, the claim recites an abstract idea.
Under Step 2A Prong Two, the claim as a whole is not integrated into a practical application, because the additional elements recited in the claim beyond the identified judicial exception include “at least one processor; and at least one memory storing instructions that, when executed by the at least one processor,” which is the use of generic computer components (at least one processor, at least one memory) to implement the abstract idea, (MPEP § 2106.05(f)), that does not serve to integrate the abstract idea into a practical application. The claim also recites a “machine learning model,” which is recited at a high-level of generality, and accordingly, is a generic computer component used to implement the abstract idea, (MPEP § 2106.05(f)), that does not serve to integrate the abstract idea into a practical application.
The claim also recites the limitation of “sending a request to validate a machine learning model,” which is a pre-processing, insignificant extra-solution activity of providing data, (MPEP § 2106.05(g)), that does not serve to integrate the abstract idea into a practical application. The claim further recites more details or specifics to the additional element of “sending,” “wherein the request comprises a data structure of the machine learning model and an indication of a current condition under which inference by the machine learning model is to be executed,” and accordingly, is merely more specific to the additional element. Therefore, the claim is directed to the abstract idea.
Finally, under Step 2B, the additional elements, taken alone or in combination, do not represent significantly more than the abstract idea itself. The additional elements recited in the claim beyond the identified judicial exception include “at least one processor; and at least one memory storing instructions that, when executed by the at least one processor,” which is the use of generic computer components (at least one processor, at least one memory) to implement the abstract idea, (MPEP § 2106.05(f)), that does not amount to significantly more than the abstract idea. The claim also recites a “machine learning model,” which is recited at a high-level of generality, and accordingly, is a generic computer component used to implement the abstract idea, (MPEP § 2106.05(f)), that does not amount to significantly more than the abstract idea.
The claim also recites the limitation of “sending a request to validate a machine learning model,” which is a well-understood, routine, and conventional activity of transmitting data over a network, (MPEP § 2106.05(d) sub II.i), that does not amount to significantly more than the abstract idea. The claim further recites more details or specifics to the additional element of “sending,” “wherein the request comprises a data structure of the machine learning model and an indication of a current condition under which inference by the machine learning model is to be executed,” and accordingly, is merely more specific to the additional element. Therefore, claim 6 is subject-matter ineligible.
Claim 7 depends directly or indirectly from claim 6. The claim further recites “perform inhibiting sending the machine learning model along with the data structure in the request to validate the machine learning model.” The plain and ordinary meaning of “inhibiting sending” is to simply send the data structure, which is the pre-processing, insignificant extra-solution activity of sending data, (MPEP § 2106.05(g)), that does not serve to integrate the abstract idea into a practical application. Also, this a well-understood, routine, and conventional activity of transmitting data over a network, (MPEP § 2106.05(d) sub II.i), that does not amount to significantly more than the abstract idea. Therefore, claim 7 is subject-matter ineligible.
Claim 8 depends directly or indirectly from claim .The claim further recites the limitation “performing the inference by the machine learning model based on the updated at least one of the data structure and the machine learning model.” The activity of “performing the inference” contains a limitation that can practically be performed in the human mind, including, for example, observations, evaluations, judgments, and opinions, and accordingly, is a mental process, (MPEP § 2106.04(a)(2) sub III), which is one of the groupings of abstract ideas. (MPEP § 2106.04(a)(2)).
Under Step 2A Prong Two, the claim further recites “receiving at least one of an update of the data structure or an update of the machine learning model in response to the request to validate the machine learning model,” and “updating the at least one of the data structure and the machine learning model based on the received update,” which are the post-processing insignificant extra-solution activity of data collection, (MPEP ¶ 2106.05(g)), that does not serve to integrate the abstract idea into a practical application. Under Step 2B, the limitation of “receiving at least one of an update of the data structure or an update of the machine learning model in response to the request to validate the machine learning model” is the well-understood, routine, and conventional activity of receiving data over a network, (MPEP § 2106.05(d) sub II.i), that does not amount to significantly more than the abstract idea. Also, the activity of “updating the at least one of the data structure and the machine learning model based on the received update,” which is the well-understood, routine, and conventional activity of storing information in memory, (MPEP § 2106.05(d) sub II.iv), that does not amount to significantly more than the abstract idea. Therefore, claim 8 is subject-matter ineligible.
Claims 9 and 10 depend directly or indirectly from claim 6. The claims recite more details or specifics to the additional element of “sending a request,” (claim 9: wherein the data structure of the machine learning model comprises an identifier of a machine learning model, and at least one of the following: static information on the machine learning model; dynamic information on the machine learning model; or secure information on the machine learning model; . . .”; claim 10: wherein at least one of the following: the condition under which the machine learning model was trained comprises at least one of the following: a network at which the machine learning model was trained; a radio parameter under which the machine learning model was trained; a radio condition under which the machine learning model was trained; or a parameter of a terminal under which the machine learning model was trained; or the structure of the training data used to train the machine learning model comprises at least one of the following: an input parameter of the machine learning model; a range of values of the input parameter in the training data; or a distribution of the values of the input parameter in the training data; or the secure information is encrypted in the data structure), and accordingly, are merely more specific to the additional element. Therefore, claims 9 and 10 are subject-matter ineligible.
Claim Rejections - 35 U.S.C. § 102
7. 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.
8. Claims 1-15 are rejected under 35 U.S.C. § 102(a)(2) as being anticipated by US Published Application 20250071029 to Soldati et al. [hereinafter Soldati].
Regarding claims 1 and 11, Soldati teaches [a]n apparatus (Soldati, Abstract, teaches “[s]ystems [(that is, an apparatus)] . . . for inter-node verification of models are disclosed”) of claim 1, and [a] method (Soldati, Abstract, teaches “methods [(that is, a method)] for inter-node verification of models are disclosed”) of claim 11, comprising:
at least one processor; and at least one memory storing instructions that, when executed by the at least one processor (Soldati ¶ 0376 teaches “a computer program including instructions which, when executed by at least one processor, causes the at least one processor to carry out the functionality of radio access node 1400 or a node (e.g., a processing node 1500) implementing one or more of the functions 1510 of the radio access node 1400”), cause the apparatus at least to perform:
receiving a request to validate a machine learning model (Soldati, Fig. 5, teaches a validation request, validation, and provide the validation result [Examiner annotations in dashed-line text boxes]:
PNG
media_image1.png
548
971
media_image1.png
Greyscale
Soldati ¶ 0044 teaches “a first network node to provide a model to a second network node [(that is, “providing to a second network node” is the second network node receiving)] together with configurations/instructions/semantics information for verifying (e.g., testing and/or validating) the model”),
wherein the request comprises a data structure of the machine learning model (Soldati ¶ 0077 teaches, regarding the first message, that “[w]eight factors for each input needed by the model, namely revealing the importance/priority of each input type with respect to the process of inference carried out by the model [(that is, “weight factors” are a data structure of the machine learning model)]”; also, Soldati ¶ 0170 teaches “the first network node saves and transmits the model as serialized file associated to a certain AI/ML framework (e.g., TensorFlow)”) and an indication of a current condition under which inference by the machine learning model is to be executed (Soldati ¶ 0072 teaches, regarding the first message, that “One or more conditions or events to be fulfilled for triggering the verification of the AIML model [(that is, the machine learning model is to be executed)] indicated by the first network node [(that is, an indication of a current condition under which inference by the machine learning model is to be executed)]”);
validating the machine learning model based on the data structure and the current condition to obtain a validation result (Soldati ¶¶ 0203-04 teaches “verifying an AIML model can comprise one or more of the following operations [(that is, validating the machine learning model)]: Verifying that the AIML model can be set up, applied, or installed by a network node, i.e., the AIML model and associated instructions or policies on how to set up [(that is, validating the machine learning model based on the data structure)], secure, apply, or install the AIML model are exhaustive, understood, and executable by the network node”; Soldati ¶ 0205 teaches “[t]esting whether the AIML model was correctly set up, secured, applied, or installed by a network node, e.g., by testing whether the AIML model (precisely) reproduces given reference outputs for given reference inputs. Testing whether the execution of the AIML model (i.e., inference) at a network node meets the given execution time requirement [(that is, validating the machine learning model based on . . . the current condition to obtain a validation result)]”); and
providing the validation result in response to the request to validate the machine learning model (Soldati ¶ 0049 teaches “the first network node may require the second network node to provide information related to verifying (testing and/or validation) or any other kind of evaluation of an AIML model provided by the first network node (for instance, upon the second network node or a third network node having re-trained/modified AIML model to the first network node) [(that is, providing the validation result in response to the request to validate the machine learning model)]. In one example, the first network node can control whether, when, and how it must be notified by the second network node about the result of verification and/or validation”), wherein
the data structure of the machine learning model comprises at least one of metadata of the machine learning model or context data of the machine learning model (Soldati ¶ 0216 teaches “the first network node 500 specifies that verifying the indicated AIML model may consist in validating the performance of the AIML model with respect to one or more hyperparameters [(that is, training process configuration settings)]of the AIML model [(that is, “hyperparameters” are the data structure of the machine learning model comprises . . . context data of the machine learning model)]”).
Regarding claims 2 and 12, Soldati teaches all of the limitations of claims 1 and 11, respectively, as described above in detail.
Soldati teaches -
wherein the instructions, when executed by the at least one processor, cause the apparatus to perform
the validating without using the machine learning model (Soldati ¶ 0191 teaches “the second network node 502 may without previous configurations/instructions from the first network node 500, run the model verification process and notify the first network node 500 of inconsistencies between the instructions/configurations/semantics relative to the inputs and/or outputs, provided prior to using the AIML model, and the actual availability of the inputs and/or outputs over the connected interfaces to the second network node 502 [(that is, the validating without using the machine learning model)]”).
Regarding claims 3 and 13, Soldati teaches all of the limitations of claims 1 and 11, respectively, as described above in detail.
Soldati teaches -
wherein the instructions, when executed by the at least one processor, further cause the apparatus to perform
providing at least one of an update of the data structure or an update of the machine learning model (Soldati ¶ 0099 teaches “[a] non-limiting example is model re-training [(that is, an update)]. That is, when the AIML model indicated by the FIRST MESSAGE is re-trained by the second network node or by a third network node [(that is, providing at least one of . . . an update of the machine learning model)]”; notably, Soldati ¶ 0164 teaches that “[t]he terms model training, model optimizing, model optimization, model updating are herein used interchangeably with the same meaning unless explicitly specified otherwise”) in response to the request to validate the machine learning model (Soldati ¶ 0200 teaches “the reference set of data samples for verifying the AIML model could be used by the second network node 502 (or by a third network node) to test or validate an AIML model provided by the first network node 500 in case the second network node 502 (or a third network node) determines to re-train the AIML model [(that is, in response to the request to validate the machine learning model)]”).
Regarding claims 4 and 14, Soldati teaches all of the limitations of claims 1 and 11, respectively, as described above in detail.
Soldati teaches -
wherein the data structure of the machine learning model comprises
an identifier of a machine learning model (Soldati ¶ 0066 teaches an “identity or an identifier of an AIML model to which the configuration for verification is applicable to or associated to [(that is, an identifier of a machine learning model)]”), and at least one of the following:
static information on the machine learning model (see below, Soldati ¶ 0198);
dynamic information on the machine learning model (see below, Soldati ¶¶ 0039-40, 0244-45); or
secure information on the machine learning model (see below, Soldati ¶ 0171);
wherein the static information comprises at least one of the following:
an indication of an architecture of the machine learning model;
a number of layers of the machine learning model;
an optimizer used to derive the machine learning model;
an indication if the machine learning model is one-sided or two-sided;
a format of the machine learning model;
an indication on a condition under which the machine learning model was trained;
an indication of training data used to train the machine learning model (Soldati ¶ 0198 teaches “the reference set of data samples for verifying the AIML model could consist of a set of reference input-output pairs, where each reference output value represents that output that is expected to obtain for the corresponding reference input data when provided to the model for verification [(that is, an indication of training data used to train the machine learning model)]”);
a structure of the training data used to train the machine learning model; or
a geographical location at which the machine learning model was trained;
the dynamic information comprises at least one of the following:
the indication on the condition under which the machine learning model was trained (Soldati ¶¶ 0244-45 teaches “If the environment in which the AIML model operates changes. Such changes may constitute of: Changes in the capabilities of the second network node 502 that may condition the functioning of the AIML model, e.g., changes in hardware capabilities or compute resources ( e.g., processing power, memory) available or allocated for the execution of the AIML model [(that is, the indication on the condition under which the machine learning was trained)]”);
the indication of the training data used to train the machine learning model (Soldati ¶¶ 0039-40 teaches that “when the model is trained, validating the model with different set of data (e.g., different from training data) provides an opportunity to further improve the model quality, which further avoids making wrong decisions taken by the machine in the real-life prediction. In this case, besides training data provided to "Model Training" function and inference data provided to "Model Inference" function, "Data Collection" should also provide validation data to "Model Training", so that the accuracy of the trained model can be guaranteed”);
the structure of the training data used to train the machine learning model; or
the geographical location at which the machine learning model was trained;
the secure information comprises at least one of the following:
a usage experience of the machine learning model (Soldati ¶ 0171 teaches “the first network node trains a model for a use case that operates on a very fast time scale and thus requires a very low execution (i.e., inference) time. In this case, a problem occurs if, for example, the hardware capabilities of the other network node or the compute resources allocated by the other network node for the execution of the model do not allow the other network node to meet the time requirement of the use case”); or
a dependence of the user experience on a hardware or a chipset or a system on chip.
Regarding claims 5 and 15, Soldati teaches all of the limitations of claims 4 and 14, respectively, as described above in detail.
Soldati teaches -
wherein at least one of the following:
the condition under which the machine learning model was trained comprises at least one of the following: a network at which the machine learning model was trained (Soldati ¶ 0145 teaches “Network Node: As used herein, a "network node" is any node that is either part of the RAN or the core network of a cellular communications network/system [(that is, a “network node” is the condition under which the machine learning model was trained comprises at least one of the following: a network at which the machine learning model was trained)]”); a radio parameter under which the machine learning model was trained; a radio condition under which the machine learning model was trained; or a parameter of a terminal under which the machine learning model was trained; or
the structure of the training data used to train the machine learning model comprises at least one of the following: an input parameter of the machine learning model; a range of values of the input parameter in the training data; or a distribution of the values of the input parameter in the training data; or the secure information is encrypted in the data structure.
Regarding claim 6, Soldati teaches [a]n apparatus (Soldati, Abstract, teaches “[s]ystems [(that is, an apparatus)] . . . for inter-node verification of models are disclosed”), comprising:
at least one processor; and at least one memory storing instructions that, when executed by the at least one processor (Soldati ¶ 0376 teaches “a computer program including instructions which, when executed by at least one processor, causes the at least one processor to carry out the functionality of radio access node 1400 or a node (e.g., a processing node 1500) implementing one or more of the functions 1510 of the radio access node 1400”), cause the apparatus at least to perform:
sending a request to validate a machine learning model (Soldati, Fig. 5, teaches a validation request, validation, and provide the validation result [Examiner annotations in dashed-line text boxes]:
PNG
media_image2.png
681
1221
media_image2.png
Greyscale
Soldati ¶ 0044 teaches “a first network node to provide a model to a second network node [(that is, sending a request to validate a machine learning model)] together with configurations/instructions/semantics information for verifying (e.g., testing and/or validating) the model”),
wherein the request comprises a data structure of the machine learning model (Soldati ¶ 0077 teaches, regarding the first message, that “[w]eight factors for each input needed by the model, namely revealing the importance/priority of each input type with respect to the process of inference carried out by the model [(that is, “weight factors” are a data structure of the machine learning model)]”; also, Soldati ¶ 0170 teaches “the first network node saves and transmits the model as serialized file associated to a certain AI/ML framework (e.g., TensorFlow)”) and an indication of a current condition under which inference by the machine learning model is to be executed (Soldati ¶ 0072 teaches, regarding the first message, that “One or more conditions or events to be fulfilled for triggering the verification of the AIML model [(that is, the machine learning model is to be executed)] indicated by the first network node [(that is, an indication of a current condition under which inference by the machine learning model is to be executed)]”);
receiving a validation result in response to the request to validate (Soldati ¶ 0049 teaches “the first network node may require the second network node to provide information related to verifying (testing and/or validation) or any other kind of evaluation of an AIML model provided by the first network node (for instance, upon the second network node or a third network node having re-trained/modified AIML model to the first network node) [(that is, receiving a validation result in response to the request to validate)]. In one example, the first network node can control whether, when, and how it must be notified by the second network node about the result of verification and/or validation”); and
deciding whether or not to perform inference by the machine learning model based on the validation result (Soldati ¶ 0044 teaches “[i]n some embodiments, the model is an Artificial Intelligence (Al) and/or Machine Learning (ML) model. This may enable a first network node, responsible for training an AIML model and providing it to a second network node, to specify whether, when, and how the model can, should, or must be verified by the second network node prior to using the model [(that is, the validation result)]. This allows the first network node to ensure that the model, for which it is responsible, is correctly set up, applied, or installed by the second network node and works as expected before the second network uses the model, for instance, for inference [(that is, deciding whether or not to perform inference by the machine learning model based on the validation result)]”; Soldati ¶ 0047 teaches “the second network node could be required to validate that the inputs required by the model can be received over the interfaces connecting the second network node to other parts of the system, according to the received instructions/configurations/semantics provided prior to using the AIML model, e.g., for inference”), wherein
the data structure of the machine learning model comprises at least one of metadata of the machine learning model or context data of the machine learning model (Soldati ¶ 0216 teaches “the first network node 500 specifies that verifying the indicated AIML model may consist in validating the performance of the AIML model with respect to one or more hyperparameters [(that is, training process configuration settings)]of the AIML model [(that is, “hyperparameters” are the data structure of the machine learning model comprises . . . context data of the machine learning model)]”).
Regarding claim 7, Soldati teaches all of the limitations of claim 6, as described above in detail.
Soldati teaches -
wherein the instructions, when executed by the at least one processor, further cause the apparatus to
perform inhibiting sending the machine learning model along with the data structure in the request to validate the machine learning model (Soldati ¶ 0062 teaches that “[i]n one embodiment of the method, the first network node may provide to the second network node [(that is, sending)], either with the FIRST MESSAGE or with a THIRD MESSAGE, a set of reference data samples which can be used to verify the AIML model. In one example the set of reference data samples can be provided with the FIRST MESSAGE as part of the configuration for verifying an AI/ML model. In another example, the provided reference data samples could be explicitly associated to one or more AIML models [(that is, the “provide . . . a set of reference data samples” is perform inhibiting sending the machine learning model along with the data structure in the request to validate the machine learning model)]”; cf. Soldati ¶ 0061, which teaches “[i]n one embodiment, the first network node may provide, either with the FIRST MESSAGE or with a THIRD MESSAGE, an AIML model to the second network node and the configurations/instructions/semantics information for verifying an AI/ML model associated to the AIML model provided by the first network node”;
[Examiner notes that the plain meaning of “inhibiting sending” is to not send, or send without, something. Accordingly, the broadest reasonable interpretation of “inhibiting sending” is being omitted from the request, which is not inconsistent with the Applicant’s disclosure (MPEP § 2111). Thus, term “inhibiting sending” covers the teachings of Soldati relating to the “providing a set of reference data samples,” which is further clarified as “inhibiting sending the machine learning model” when the provided reference data samples provided to the second node of Soldati are “explicitly associated to one or more AIML models” (Soldati ¶ 0062))].
Regarding claim 8, Soldati teaches all of the limitations of claim 6, as described above in detail.
Soldati teaches -
wherein the instructions, when executed by the at least one processor, further cause the apparatus to perform
receiving at least one of an update of the data structure or an update of the machine learning model (Soldati ¶ 0099 teaches “[a] non-limiting example is model re-training [(that is, an update)]. That is, when the AIML model indicated by the FIRST MESSAGE is re-trained by the second network node or by a third network node [(that is, receiving at least one of . . . an update of the machine learning model)]”; notably, Soldati ¶ 0164 teaches that “[t]he terms model training, model optimizing, model optimization, model updating are herein used interchangeably with the same meaning unless explicitly specified otherwise”) in response to the request to validate the machine learning model (Soldati ¶ 0200 teaches “the reference set of data samples for verifying the AIML model could be used by the second network node 502 (or by a third network node) to test or validate an AIML model provided by the first network node 500 in case the second network node 502 (or a third network node) determines to re-train the AIML model [(that is, in response to the request to validate the machine learning model)]”);
updating the at least one of the data structure and the machine learning model based on the received update (Soldati, Fig. 12, teaches model updating [Examiner annotations in dashed-line text boxes]:
PNG
media_image3.png
639
1101
media_image3.png
Greyscale
Soldati ¶ 0024 teaches “(FFS) Model Performance Feedback: Applied if certain information derived from Model Inference function is suitable for improvement of the AI/ML model trained in Model Training function. Feedback from Actor or other network entities (via Data Collection function) may be needed at Model Inference function to create Model Performance Feedback”; Soldati ¶ 0029 teaches “Model Deployment/Update: Used to initially deploy a trained, validated, and tested AI/ML model to the Model Inference function or to deliver an updated model to the Model Inference function”; also, notably, Soldati ¶ 0164 teaches “[t]he terms model training, model optimizing, model optimization, model updating are herein used interchangeably with the same meaning unless explicitly specified otherwise [(that is, updating the at least one of the data structure and the machine learning model based on the received update)]”); and
performing the inference by the machine learning model based on the updated at least one of the data structure and the machine learning model Soldati ¶ 0029 & Fig. 12 teaches “Model Deployment/Update: Used to initially deploy a trained, validated, and tested AI/ML model to the Model Inference function or to deliver an updated model to the Model Inference function [(that is, performing the inference by the machine learning model based on the updated at least one of the data structure and the machine learning model)]”.
Regarding claim 9, Soldati teaches all of the limitations of claim 6, as described above in detail.
Soldati teaches -
wherein the data structure of the machine learning model comprises
an identifier of a machine learning model (Soldati ¶ 0066 teaches an “identity or an identifier of an AIML model to which the configuration for verification is applicable to or associated to [(that is, an identifier of a machine learning model)]”), and at least one of the following:
static information on the machine learning model (see below, Soldati ¶ 0198);
dynamic information on the machine learning model (see below, Soldati ¶¶ 0039-40, 0244-45); or
secure information on the machine learning model (see below, Soldati ¶ 0171); wherein
the static information comprises at least one of the following:
an indication of an architecture of the machine learning model;
a number of layers of the machine learning model;
an optimizer used to derive the machine learning model;
an indication if the machine learning model is one-sided or two-sided;
a format of the machine learning model;
an indication on a condition under which the machine learning model was trained;
an indication of training data used to train the machine learning model (Soldati ¶ 0198 teaches “the reference set of data samples for verifying the AIML model could consist of a set of reference input-output pairs, where each reference output value represents that output that is expected to obtain for the corresponding reference input data when provided to the model for verification [(that is, an indication of training data used to train the machine learning model)]”);
a structure of the training data used to train the machine learning model; or
a geographical location at which the machine learning model was trained;
the dynamic information comprises at least one of the following:
the indication on the condition under which the machine learning model was trained (Soldati ¶¶ 0244-45 teaches “If the environment in which the AIML model operates changes. Such changes may constitute of: Changes in the capabilities of the second network node 502 that may condition the functioning of the AIML model, e.g., changes in hardware capabilities or compute resources ( e.g., processing power, memory) available or allocated for the execution of the AIML model [(that is, the indication on the condition under which the machine learning was trained)]”);
the indication of the training data used to train the machine learning model (Soldati ¶¶ 0039-40 teaches that “when the model is trained, validating the model with different set of data (e.g., different from training data) provides an opportunity to further improve the model quality, which further avoids making wrong decisions taken by the machine in the real-life prediction. In this case, besides training data provided to "Model Training" function and inference data provided to "Model Inference" function, "Data Collection" should also provide validation data to "Model Training", so that the accuracy of the trained model can be guaranteed”);
the structure of the training data used to train the machine learning model; or
the geographical location at which the machine learning model was trained;
the secure information comprises at least one of the following:
a usage experience of the machine learning model (Soldati ¶ 0171 teaches “the first network node trains a model for a use case that operates on a very fast time scale and thus requires a very low execution (i.e., inference) time. In this case, a problem occurs if, for example, the hardware capabilities of the other network node or the compute resources allocated by the other network node for the execution of the model do not allow the other network node to meet the time requirement of the use case”); or
a dependence of the user experience on a hardware or a chipset or a system on chip.
Regarding claim 10, Soldati teaches all of the limitations of claim 9, as described above in detail.
Soldati teaches -
wherein at least one of the following:
the condition under which the machine learning model was trained comprises at least one of the following: a network at which the machine learning model was trained (Soldati ¶ 0145 teaches “Network Node: As used herein, a "network node" is any node that is either part of the RAN or the core network of a cellular communications network/system [(that is, a “network node” is the condition under which the machine learning model was trained comprises at least one of the following: a network at which the machine learning model was trained)]”); a radio parameter under which the machine learning model was trained; a radio condition under which the machine learning model was trained; or a parameter of a terminal under which the machine learning model was trained; or
the structure of the training data used to train the machine learning model comprises at least one of the following: an input parameter of the machine learning model; a range of values of the input parameter in the training data; or a distribution of the values of the input parameter in the training data; or
the secure information is encrypted in the data structure.
Conclusion
9. The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
(US Published Application 20210274361 to Gupta et al.) teaches machine learning deployment in radio access networks. A machine learning deployment pipeline can comprise a machine learning model design platform, a network automation platform, and a radio access network.
(Balaji Roghothaman, “Training, Testing, and Validation Challenges for Next Generation AI/ML-Based Intelligent Wireless Networks,” IEEE (2021)) teaches a network where artificial intelligence and deep learning are native and pervasive. The disaggregation of the network and the evolution towards an open RAN architecture has provided a framework for more innovative approaches.
10. Any inquiry concerning this communication or earlier communications from the Examiner should be directed to KEVIN L. SMITH whose telephone number is (571) 272-5964. Normally, the Examiner is available on Monday-Thursday 0730-1730.
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, KAKALI CHAKI can be reached on 571-272-3719. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/K.L.S./
Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122
Prosecution Insights