The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Priority
Receipt is acknowledged of certified copies of papers required by 37 CFR 1.55.
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph:
(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and
(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function.
Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function.
Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action.
This application includes one or more claim limitations that do not use the word “means,” but are nonetheless being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, because the claim limitation(s) uses a generic placeholder that is coupled with functional language without reciting sufficient structure to perform the recited function and the generic placeholder is not preceded by a structural modifier. Such claim limitation(s) is/are:
Claim No.
Limitation
1-9
at least one module for updating the at least one ML model (structure not disclosed)
1-9
a scheduler for managing executions of FL tasks in the cellular network (structure not disclosed)
Because this/these claim limitation(s) is/are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, it/they is/are being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof.
If applicant does not intend to have this/these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitation(s) recite(s) sufficient structure to perform the claimed function so as to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph.
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 1-9 rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
Claim limitations “at least one module for updating the at least one ML”, and “a scheduler for managing executions of FL tasks in the cellular network” invoke 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. However, the written description fails to disclose the corresponding structure, material, or acts for performing the entire claimed function and to clearly link the structure, material, or acts to the function. The various disclosures provided in the specification are qualified with "may" or as one of many possible embodiments. Therefore, the claim is indefinite and is rejected under 35 U.S.C. 112(a) or pre-AIA 35 U.S.C. 112, second paragraph.
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
Claims 1-9 rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim limitations “at least one module for updating the at least one ML”, and “a scheduler for managing executions of FL tasks in the cellular network” invoke 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. However, the written description fails to disclose the corresponding structure, material, or acts for performing the entire claimed function and to clearly link the structure, material, or acts to the function. The various disclosures provided in the specification are qualified with "may" or as one of many possible embodiments. Therefore, the claim is indefinite and is rejected under 35 U.S.C. 112(b) or pre-AIA 35 U.S.C. 112, second paragraph.
Applicant may:
(a) Amend the claim so that the claim limitation will no longer be interpreted as a limitation under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph;
(b) Amend the written description of the specification such that it expressly recites what structure, material, or acts perform the entire claimed function, without introducing any new matter (35 U.S.C. 132(a)); or
(c) Amend the written description of the specification such that it clearly links the structure, material, or acts disclosed therein to the function recited in the claim, without introducing any new matter (35 U.S.C. 132(a)).
If applicant is of the opinion that the written description of the specification already implicitly or inherently discloses the corresponding structure, material, or acts and clearly links them to the function so that one of ordinary skill in the art would recognize what structure, material, or acts perform the claimed function, applicant should clarify the record by either:
(a) Amending the written description of the specification such that it expressly recites the corresponding structure, material, or acts for performing the claimed function and clearly links or associates the structure, material, or acts to the claimed function, without introducing any new matter (35 U.S.C. 132(a)); or
(b) Stating on the record what the corresponding structure, material, or acts, which are implicitly or inherently set forth in the written description of the specification, perform the claimed function. For more information, see 37 CFR 1.75(d) and MPEP §§ 608.01(o) and 2181.
Claim Rejections - 35 USC § 101
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.
Claims 1-9 and 13-19 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim(s) recite(s) a mathematical concept; "at least one ML model for training" and various mathematical calculations involving said ML model in claims 1-9; and mental processes in claims 1-9 and 13-20.
Regarding claim 1:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 1 is directed to software per se.
Step 1: no
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?",
At least one ML model for training
is directed to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation)
Select a set of user equipments (UEs), from a plurality of subscribing UEs, as being suitable for the FL task,
is directed to an abstract idea (i.e., a mental process, e.g., evaluation).
Determine a clustering policy for the FL task, which specifies how to group the selected UEs into a plurality of clusters
is directed to an abstract idea (i.e., a mental process, e.g., judgement).
Step 2A prong 1: yes
Per step 2A prong 2, "Does the claim recite additional elements that integrate the judicial exemption into a practical application?"
A parameter server for training machine learning (ML) models in a cellular network, the parameter server comprising (well-understood, routine, and conventional generic computer, see MPEP 2106.05(f))
At least one module for updating the at least one ML model as a response to a federated learning (FL) task being executed (well-understood, routine, and conventional generic computer, see MPEP 2106.05(f))
A scheduler for managing executions of FL tasks in the cellular network, wherein for each FL task, the scheduler is configured to (well-understood, routine, and conventional generic computer, see MPEP 2106.05(f))
Instruct each cluster of the plurality of clusters to perform the FL task (insignificant extra-solution activity of mere data output, see MPEP 2106.05(g))
Step 2A prong 2: no
Per step 2B, does the claim recite additional elements that amount to significantly more than the judicial exception, the claims do not include additional elements that are sufficient to amount to more than the judicial exception. Specifically, the claimed inventions simply append well-understood, routine, and conventional activities previously known to the industry (see analysis under prong 2), both when viewed independently and as an ordered combination, specified at a high level of generality, to the judicial exception (e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine, and conventional activities previously known to the industry).
Step 2B: no
Regarding claim 2:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 2 is directed to software per se.
Step 1: yes
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?"
The parameter server of claim 1
is directed to an abstract idea (i.e., a mental process, e.g., evaluation) and to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation).
Determine, using the stored information on the plurality of subscribing UEs, whether the at least one condition is satisfied by the plurality of subscribing UEs
is directed to an abstract idea (i.e., a mental process, e.g., evaluation).
Step 2A prong 1: yes
Per step 2A prong 2, "Does the claim recite additional elements that integrate the judicial exemption into a practical application?"
Wherein the scheduler is further configured to: store information on the plurality of subscribing UEs that have each subscribed to perform at least one FL task (well-understood, routine, and conventional generic computer, see MPEP 2106.05(f))
Receive, from the at least one module, a request to run or re-run an FL task with respect to a specific ML model the request specifying at least one condition to be satisfied by UEs performing the requested FL task (insignificant extra-solution activity of mere data gathering, see MPEP 2106.05(g))
Transmit, to the least one module, a response indicating whether the request is granted based on the determination (insignificant extra-solution activity of mere data output, see MPEP 2106.05(g))
Step 2A prong 2: no
Per step 2B, does the claim recite additional elements that amount to significantly more than the judicial exception, the claims do not include additional elements that are sufficient to amount to more than the judicial exception. Specifically, the claimed inventions simply append well-understood, routine, and conventional activities previously known to the industry (see analysis under prong 2), both when viewed independently and as an ordered combination, specified at a high level of generality, to the judicial exception (e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine, and conventional activities previously known to the industry).
Step 2B: no
Regarding claim 3:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 3 is directed to software per se.
Step 1: no
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?"
The parameter server of claim 2
is directed to an abstract idea (i.e., a mental process, e.g., evaluation) and to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation).
wherein the at least one condition specifies any one or more of a quality of service (QoS) profile, a minimum number of UEs required to perform the FL task, and a UE hardware capacity requirement for performing the FL task
is directed to an abstract idea (i.e., a mental process, e.g., evaluation).
Step 2A prong 1: yes
Claim 3 does not integrate into a practical application (2A prong 2) or add any additional elements (2B).
Regarding claim 4:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 4 is directed to software per se.
Step 1: no
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?"
The parameter server of claim 3
is directed to an abstract idea (i.e., a mental process, e.g., evaluation) and to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation).
Before a training iteration begins, the scheduler is configured to: check UE status messages received from the UEs in the plurality of clusters to determine whether the UEs are still able to perform the training iteration of the FL task
is directed to an abstract idea (i.e., a mental process, e.g., observation).
Step 2A prong 1: yes
Per step 2A prong 2, "Does the claim recite additional elements that integrate the judicial exemption into a practical application?"
Wherein the FL task comprises a plurality of training iterations (well-understood, routine, and conventional generic computer, see MPEP 2106.05(f))
Instruct, responsive to determining at least one UE is unable to perform the training iteration of the FL task, a coordinator to re-group the UEs into a plurality of clusters according to the clustering policy and the UE status messages (insignificant extra-solution activity of mere data output, see MPEP 2106.05(g))
Step 2A prong 2: yes
Per step 2B, does the claim recite additional elements that amount to significantly more than the judicial exception, the claims do not include additional elements that are sufficient to amount to more than the judicial exception. Specifically, the claimed inventions simply append well-understood, routine, and conventional activities previously known to the industry (see analysis under prong 2), both when viewed independently and as an ordered combination, specified at a high level of generality, to the judicial exception (e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine, and conventional activities previously known to the industry).
Step 2B: no
Regarding claim 5: Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 5 is directed to software per se.
Step 1: no
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?"
The parameter server of claim 4
is directed to an abstract idea (i.e., a mental process, e.g., evaluation) and to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation).
update the ML model corresponding to the FL task that has been performed by the UE s using the first and second set
is directed to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation)
wherein the first set of parameters is generated based on a training of a local version of the ML model corresponding to the FL task using at least one training data item stored on the UE, and
is directed to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation)
wherein the second set of parameters is generated based on a training of a local version of the ML model corresponding to the FL task using at least one training data item generated by the UE and at least one generated coded training data item received from other UEs in the cluster
is directed to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation)
Step 2A prong 1: yes
Per step 2A prong 2, "Does the claim recite additional elements that integrate the judicial exemption into a practical application?"
wherein at least one module of the parameter server is configured to: receive, from a UE via the scheduler, a first set of parameters and a second set of parameters (insignificant extra-solution activity of mere data gathering, see MPEP 2106.05(g))
Step 2A prong 2: no
Per step 2B, does the claim recite additional elements that amount to significantly more than the judicial exception, the claims do not include additional elements that are sufficient to amount to more than the judicial exception. Specifically, the claimed inventions simply append well-understood, routine, and conventional activities previously known to the industry (see analysis under prong 2), both when viewed independently and as an ordered combination, specified at a high level of generality, to the judicial exception (e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine, and conventional activities previously known to the industry).
Step 2B: no.
Regarding claim 6:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 6 is directed to software per se.
Step 1: yes
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?"
The parameter server of claim 5
is directed to an abstract idea (i.e., a mental process, e.g., evaluation) and to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation).
Wherein when the first set of parameters has been received from a pre-defined minimum number of UEs from the plurality of clusters within a pre-defined time period, the at least one module is configured to update the ML model by: aggregating the first set of parameters received from the UEs in the plurality of clusters, and updating the ML model using the aggregated first set of parameters
is directed to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation)
Step 2A prong 1: yes
Claim 6 does not integrate into a practical application (2A prong 2) or add any additional elements (2B).
Regarding claim 7:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 7 is directed to software.
Step 1: no
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?"
The parameter server of claim 5
is directed to an abstract idea (i.e., a mental process, e.g., evaluation) and to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation).
wherein when the first set of parameters has been received from fewer than a pre-defined minimum number of UEs from the plurality of clusters within a pre-defined time period, the at least one module is configured to update the ML model by: aggregating the first set of parameters received from the UEs in the plurality of clusters
is directed to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation)
aggregating a random selection of the second set of parameters received from the UEs
is directed to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation)
and updating the ML model using the aggregated first set of parameters and the aggregated random selection of the second set of parameters
is directed to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation)
Step 2A prong 1: yes
Claim 7 does not integrate into a practical application (2A prong 2) or add any additional elements (2B).
Regarding claim 8:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 8 is directed to software per se.
Step 1: no
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?"
The parameter server of claim 7
is directed to an abstract idea (i.e., a mental process, e.g., evaluation) and to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation).
Step 2A prong 1: yes
Per step 2A prong 2, "Does the claim recite additional elements that integrate the judicial exemption into a practical application?"
Wherein when the second set of parameters has been received from fewer than a pre-defined minimum number of UEs in a cluster within a pre-defined time period, the at least one module is configured to terminate the updating of the ML model. (insignificant extra-solution activity of mere data output, see MPEP 2106.05(g))
Step 2A prong 2: no
Per step 2B, does the claim recite additional elements that amount to significantly more than the judicial exception, the claims do not include additional elements that are sufficient to amount to more than the judicial exception. Specifically, the claimed inventions simply append well-understood, routine, and conventional activities previously known to the industry (see analysis under prong 2), both when viewed independently and as an ordered combination, specified at a high level of generality, to the judicial exception (e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine, and conventional activities previously known to the industry).
Step 2B: no
Regarding claim 9:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 9 is directed to software per se.
Step 1: no
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?"
The parameter server of claim 1
is directed to an abstract idea (i.e., a mental process, e.g., evaluation) and to an abstract idea (i.e. a mathematical concept, e.g. a mathematical calculation).
Step 2A prong 1: yes
Per step 2A prong 2, "Does the claim recite additional elements that integrate the judicial exemption into a practical application?"
the cellular network is an open radio-access network (ORAN), (well-understood, routine, and conventional generic computer, see MPEP 2106.05(f)) the parameter server is a service management and orchestration (SMO) platform comprising a non-real time radio intelligent controller (non-RT-RIC), (well-understood, routine, and conventional generic computer, see MPEP 2106.05(f)) the at least one module for updating the at least one ML model is a software application (rAPP) configured to run on the non-RT-RIC, (well-understood, routine, and conventional generic computer, see MPEP 2106.05(f)) and the scheduler is a software application configured to run on the non-RT-RIC. (well-understood, routine, and conventional generic computer, see MPEP 2106.05(f)) Step 2A prong 2: no.
Per step 2B, does the claim recite additional elements that amount to significantly more than the judicial exception, the claims do not include additional elements that are sufficient to amount to more than the judicial exception. Specifically, the claimed inventions simply append well-understood, routine, and conventional activities previously known to the industry (see analysis under prong 2), both when viewed independently and as an ordered combination, specified at a high level of generality, to the judicial exception (e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine, and conventional activities previously known to the industry).
Step 2B: no
Regarding claim 13:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim directed to a process, machine, manufacture, or composition of matter?”, claim 13 is directed to a process.
Step 1: yes
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?",
selecting a set of subscribing user equipments (UEs) from a plurality of subscribing UEs in the cellular network as being suitable for performing an FL task, where each subscribing UE has subscribed to perform at least one FL task;
is directed to an abstract idea (i.e., a mental process, e.g. evaluation).
Determining a clustering policy for the FL task, which specifies how to group the selected set of subscribing UEs into a plurality of clusters;
is directed to an abstract idea (i.e., a mental process, e.g., evaluation).
Step 2A prong 1: yes
Per step 2A prong 2, "Does the claim recite additional elements that integrate the judicial exemption into a practical application?"
A method performed by a parameter server for training machine learning (ML) models using federated learning (FL) in a cellular network, the method comprising (well-understood, routine, and conventional generic computer, see MPEP 2106.05(f))
instructing each cluster of the plurality of clusters of subscribing UEs to perform the FL task. (insignificant extra-solution activity of mere data output, see MPEP 2106.05(g))
Step 2A prong 2: no
Per step 2B, does the claim recite additional elements that amount to significantly more than the judicial exception, the claims do not include additional elements that are sufficient to amount to more than the judicial exception. Specifically, the claimed inventions simply append well-understood, routine, and conventional activities previously known to the industry (see analysis under prong 2), both when viewed independently and as an ordered combination, specified at a high level of generality, to the judicial exception (e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine, and conventional activities previously known to the industry).
Step 2B: no
Regarding claim 14:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 14 is directed to a process.
Step 1: yes.
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?"
The method of claim 13
is directed to an abstract idea (i.e., a mental process, e.g., evaluation. See analysis of claim 13.)
Step 2A prong 1: yes
Per step 2A prong 2, "Does the claim recite additional elements that integrate the judicial exemption into a practical application?"
further comprising: receiving a request to run or re-run an AFL task with respect to a specific ML model (insignificant extra-solution activity of mere data gathering, see MPEP 2106.05(g))
Step 2A prong 2: no.
Per step 2B, does the claim recite additional elements that amount to significantly more than the judicial exception, the claims do not include additional elements that are sufficient to amount to more than the judicial exception. Specifically, the claimed inventions simply append well-understood, routine, and conventional activities previously known to the industry (see analysis under prong 2), both when viewed independently and as an ordered combination, specified at a high level of generality, to the judicial exception (e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine, and conventional activities previously known to the industry).
Step 2B: no
Regarding claim 15:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 15 is directed to software per se.
Step 1: no
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?"
The method of claim 14
is directed to an abstract idea (i.e., a mental process, e.g., evaluation. See analysis of claim 13.
Step 2A prong 1: yes
Per step 2A prong 2, "Does the claim recite additional elements that integrate the judicial exemption into a practical application?"
further comprising: transmitting, to a coordinator for coordinating the execution of FL tasks by the UEs in each cluster, a request to determine a per-cluster data coding optimization policy to be used by each UE (insignificant extra-solution activity of mere data output, see MPEP 2106.05(g))
Step 2A prong 2: no
Per step 2B, does the claim recite additional elements that amount to significantly more than the judicial exception, the claims do not include additional elements that are sufficient to amount to more than the judicial exception. Specifically, the claimed inventions simply append well-understood, routine, and conventional activities previously known to the industry (see analysis under prong 2), both when viewed independently and as an ordered combination, specified at a high level of generality, to the judicial exception (e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine, and conventional activities previously known to the industry).
Step 2B: no
Regarding claim 16:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 16 is directed to a software per se.
Step 1: no
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?"
Grouping the set of UEs into a plurality of clusters based on the clustering policy, wherein the clustering policy specifies how to group the set of UEs into the plurality of
clusters, and wherein the set of UEs is selected from a plurality of UEs in the cellular network as being suitable for performing the FL task.
is directed to an abstract idea (i.e., a mental process, e.g., evaluation).
Step 2A prong 1: yes
Per step 2A prong 2, "Does the claim recite additional elements that integrate the judicial exemption into a practical application?"
A method performed by a coordinator for training machine learning (ML) models in a cellular network (well-understood, routine, and conventional generic computer, see MPEP 2106.05(f))
the method comprising: receiving, from a parameter server, a clustering policy for a federated learning (FL) task (insignificant extra-solution activity of mere data gathering, see MPEP 2106.05(g))
Step 2A prong 2: no
Per step 2B, does the claim recite additional elements that amount to significantly more than the judicial exception, the claims do not include additional elements that are sufficient to amount to more than the judicial exception. Specifically, the claimed inventions simply append well-understood, routine, and conventional activities previously known to the industry (see analysis under prong 2), both when viewed independently and as an ordered combination specified at a high level of generality, to the judicial exception (e.g., a claim to an abstract idea requiring no more than a generic computer to
perform generic computer functions that are well-understood, routine, and conventional activities previously known to the industry).
Step 2B: no
Regarding claim 17:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 17 is directed to software per se.
Step 1: no
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?"
The method of claim 16
is directed to an abstract idea (i.e., a mental process, e.g., evaluation. See analysis of claim 16.)
Determining a per-cluster data coding optimization policy to be used by each UE in the plurality of clusters when performing the FL task, wherein the per-cluster data coding optimization policy defines how UEs within each cluster transmit data.
is directed to an abstract idea (i.e., a mental process, e.g., judgement).
Claim 17 does not integrate into a practical application (2A prong 2) or add any additional elements (2B).
Regarding claim 18:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 19 is directed to software per se.
Step 1: no
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?"
The method of claim 17
is directed to an abstract idea (i.e., a mental process, e.g., evaluation. See analysis of claim 16.)
further comprising: periodically re-grouping, after a pre-defined time period, the set of UEs into a plurality of clusters based on the clustering policy, while the FL task is being executed
is directed to an abstract idea (i.e., a mental process, e.g., evaluation).
and re-determining, after the re-grouping, a per-cluster data coding optimization policy to be used by each UE in the plurality of clusters when performing the FL task.
is directed to an abstract idea (i.e., a mental process, e.g., judgement).
Claim 17 does not integrate into a practical application (2A prong 2) or add any additional elements (2B).
Regarding claim 19:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 19 is directed to software per se.
Step 1: no
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?"
The method of claim 16
is directed to an abstract idea (i.e., a mental process, e.g., evaluation. See analysis of claim 16.)
Step 2A prong 1: yes
Per step 2A prong 2, "Does the claim recite additional elements that integrate the judicial exemption into a practical application?"
further comprising: transmitting, to the parameter server, information on which UEs are in each cluster for the FL task, so that the parameter server is able to instruct the UEs in each cluster to perform the FL task. (insignificant extra-solution activity of mere data output, see MPEP 2106.05(g))
Step 2A prong 2: no
Per step 2B, does the claim recite additional elements that amount to significantly more than the judicial exception, the claims do not include additional elements that are sufficient to amount to more than the judicial exception. Specifically, the claimed inventions simply append well-understood, routine, and conventional activities previously known to the industry (see analysis under prong 2), both when viewed independently and as an ordered combination, specified at a high level of generality, to the judicial exception (e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine, and conventional activities previously known to the industry).
Step 2B: no
Regarding claim 20:
Per step 1 of the Subject Matter Eligibility Test for Products and Processes, “is the claim to a process, machine, manufacture, or composition of matter?”, claim 20 is directed to software per se.
Step 1: no
Per step 2A prong 1, "Does the claim recite an abstract idea, law of nature, or natural phenomenon?"
The method of claim 16
is directed to an abstract idea (i.e., a mental process, e.g., evaluation. See analysis of claim 16.)
Step 2A prong 1: yes
Per step 2A prong 2, "Does the claim recite additional elements that integrate the judicial exemption into a practical application?"
wherein the coordinator is a software application configured to run on a near-real time radio intelligent controller (near-RT-RIC) which controls the nodes of the cellular network (well-understood, routine, and conventional generic computer, see MPEP 2106.05(f))
Per step 2B, does the claim recite additional elements that amount to significantly more than the judicial exception, the claims do not include additional elements that are sufficient to amount to more than the judicial exception. Specifically, the claimed inventions simply append well-understood, routine, and conventional activities previously known to the industry (see analysis under prong 2), both when viewed independently and as an ordered combination, specified at a high level of generality, to the judicial exception (e.g., a claim to an abstract idea requiring no more than a generic computer to perform generic computer functions that are well-understood, routine, and conventional activities previously known to the industry).
Step 2B: no
Allowable Subject Matter
Claims 7 and 8 would be allowable if rewritten to overcome the rejection(s) under 35 U.S.C. 101 and 35 U.S.C. 112(b), 2nd paragraph, set forth in this Office action and to include all of the limitations of the base claim and any intervening claims.
The following is an examiner’s statement of reasons for allowance.
Claim 7 is directed towards the parameter server of a system for federated learning which involves:
A parameter server grouping client devices into clusters
Each client device calculating both parameters for its own training data, and a combination of parameters reached for each client in the cluster
When not enough clients are able to report parameters to the parameter server, combining individual client parameters with the combination parameters
(i.e. The parameter server of claim 5, wherein when the first set of parameters has been received from fewer than a pre-defined minimum number of UEs from the plurality of clusters within a pre-defined time period, the at least one module is configured to update the ML model by: aggregating the first set of parameters received from the UEs in the plurality of clusters; aggregating a random selection of the second set of parameters received from the UEs; and updating the ML model using the aggregated first set of parameters and the aggregated random selection of the second set of parameters in claim 7).
The closest art is Hosseinalipour, who teaches grouping client devices into clusters and having each client generate a combination of outputs in its cluster (on page 5 of Hosseinalipour, “The nodes engage in a cooperative scheme facilitated by D2D communications to realize the consensus/average of their local model parameters”). Hosseinalipour also teaches updating the model with the combination of outputs from a cluster (i.e., aggregating a random selection of the second set of parameters received from the UEs) (on page 5 of Hosseinalipour, “The parent node then samples parameters of one of the children and scales it by the number of children to calculate an approximate sum of the children nodes’ parameters”). However, neither Hosseinalipour or other art Bonawitz and Chai teach using both the cluster-level combinations and the individual models from each device (i.e. when the first set of parameters has been received from fewer than a pre-defined minimum number of UEs from the plurality of clusters within a pre-defined time period, the at least one module is configured to update the ML model by: aggregating the first set of parameters received from the UEs in the plurality of clusters […] and updating the ML model using the aggregated first set of parameters and the aggregated random selection of the second set of parameters), in combination with the remaining features and elements of the claimed invention.
Claim 8 would be allowable because it is dependent on claim 7.
Any comments considered necessary by applicant must be submitted no later than the payment of the issue fee and, to avoid processing delays, should preferably accompany the issue fee. Such submissions should be clearly labeled “Comments on Statement of Reasons for Allowance”.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
Claim(s) 10 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Hosseinalipour et al. (Hosseinalipour S, Azam SS, Brinton CG, Michelusi N, Aggarwal V, Love DJ, Dai H. Multi-stage hybrid federated learning over large-scale D2D-enabled fog networks. IEEE/ACM transactions on networking. 2022 Feb 4;30(4):1569-84. Henceforth referred to as Hosseinalipour).
Hosseinalipour teaches
A user equipment (UE) for training machine learning (ML) models in a cellular network, the UE comprising: (on page 3 of Hosseinalipour, “The parameters [for training machine learning (ML) models] of the end devices are carried through multiple layers of the network consisting of base stations (BSs), road-sized units (RSUs) [A user equipment (UE)], unmanned aerial vehicles (UAVs), high altitude platforms (HAPs), edge servers, and cloud servers, before reaching the main server. Devices located at different layers of the network can engage in direct communications via mobile-mobile (M2M), vehicle-vehicle (V2V), UAV-UAV (U2U), inter-edge, and inter-cloud links [a cellular network].” a storage storing a plurality of training data items; and (on page 4 of Hosseinalipour, “Each end device [a storage] n is associated with a dataset 𝒟n [a plurality of training data items]. Each element di ∈ 𝒟n of a dataset, called a training sample, is represented via a feature vector xi and a label yj for the ML task of interest)
at least one processor coupled to the storage and configured to: receive a data coding optimization policy, (on page 11 of Hosseinalipour,
PNG
media_image1.png
428
480
media_image1.png
Greyscale
.
“Number of D2D rounds for the cluster” is a data coding optimization policy; it is calculated by the parent node and applies to the entire cluster of child nodes [receive].)
receive, from a parameter server, instructions to perform a federated learning (FL) task with respect to a ML model, (on page 4 of Hosseinalipour, “To achieve this in a distributed manner, training is conducted through consecutive global iterations. At the start of global iteration k out of 𝒩, the main server [from a parameter server] possesses a parameter vector [instructions to perform a federated learning (FL) task with respect to a ML model], which propagates downstream [receive] through the hierarchy to the edge devices.”)
generate, for at least one training data item in the storage, a coded training data item, based on the received data coding optimization policy, and (on page 6 of Hosseinalipour, “Formally, during global iteration k, each node n ∈ 𝒞(k) engages in the following rounds [based on the received data coding optimization policy] of iterative updates for
t = 0, . . . ,θ(k )C -- 1:
PNG
media_image2.png
48
304
media_image2.png
Greyscale
.where z(0)b = z(k)n corresponds to node n’s initial parameter [at least one training data item in the storage], and
PNG
media_image3.png
28
40
media_image3.png
Greyscale
[a coded training data item] denotes the parameter after the D2D consensus process.”
transmit the at least one generated coded training data item to a cluster of UEs or to a node connected to the UEs in the cluster (on page 6 of Hosseinalipour, “Once D2D communications are finished in layer L|ℒ|, each parent node n ∈ 𝒩|ℒ|-1 of a cluster that operated in LUT mode selects a cluster head n’ , among its children 𝒬(k)(n) in layer L|ℒ|. This child n’ uploads [transmit] its parameter vector ŵn’(k) [the at least one generated coded training data item] to the parent node [a node connected to the UEs in the cluster].”)
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-4, 13, and 14 are rejected under 35 U.S.C. 103 as being unpatentable over
Bonawitz et al. (Bonawitz K, Eichner H, Grieskamp W, Huba D, Ingerman A, Ivanov V, Kiddon C, Konečný J, Mazzocchi S, McMahan B, Van Overveldt T. Towards federated learning at scale: System design. Proceedings of machine learning and systems. 2019 Apr 15;1:374-88. Henceforth referred to as Bonawitz.)
in view of Chai et al. (Chai Z, Ali A, Zawad S, Truex S, Anwar A, Baracaldo N, Zhou Y, Ludwig H, Yan F, Cheng Y. Tifl: A tier-based federated learning system. In Proceedings of the 29th international symposium on high-performance parallel and distributed computing 2020 Jun 23 (pp. 125-136). Henceforth referred to as Chai.).
Regarding claim 1 and analogous claim 13:
Bonawitz teaches
A parameter server for training machine learning models in a cellular network, the parameter server comprising: (on page 2 of Bonawitz, “The participants in the protocol are devices (currently Android phones) [in a cellular network] and the FL server [a parameter server], which is a cloud-based distributed service. Devices announced to the server that they are ready to run an FL task for a given Fl population. An FL population is specified by a globally unique name which identifies the learning problem, or application, which is worked on. An FL task is a specific computation for an FL population, such as training to be performed with given hyperparameters [for training machine learning models], or evaluation of trained models on local device data”.
at least one ML model for training: (on page 3 of Bonawitz, “If enough devices report in time, the round will be successfully completed and the server will update its global model [at least one ML model for training], otherwise the round is abandoned”)
at least one module for updating the at least one ML model as a response to a federated learning task being executed; (on page 3 of Bonawitz, “The server waits for the participating devices to report updates. As updates are retrieved, the server aggregates them using Federated Averaging and instructs the reporting devices when to reconnect. If enough devices report in time [as a response to a federated learning task being executed], the round will be successfully completed and the server will update its global model [module for updating the at least one ML model], otherwise the round is abandoned”)
and a scheduler for managing executions of FL tasks in the cellular network, wherein for each FL task, the scheduler is configured to: select a set of user equipments (UEs), from a plurality of subscribing UEs, as being suitable for performing the FL task, (on page 5 of Bonawitz, “Selectors [a scheduler for managing executions of FL tasks in the cellular network] are responsible for accepting and forwarding device connections. They periodically receive information from the Coordinator about how many devices are needed for each FL population [as being suitable for performing the FL task], which they use to make local decisions about whether or not to accept each device. [select a set of user equipments (UEs) from a plurality of subscribing UEs]”)
Bonawitz fails to teach
determine a clustering policy for the FL task, which specifies how to group the selected set of UEs into a plurality of clusters,
and instruct each cluster of the plurality of clusters to perform the FL task
Chai teaches
determine a clustering policy for the FL task, which specifies how to group the selected set of UEs into a plurality of clusters, (on page 4 of Chai, ”In TIFL, the first step is to collect the latency metrics of all the available clients through a lightweight profiling as detailed in section 4.2. The profiled data is further utilized by our tiering algorithm [a clustering policy for the FL task, which specifies how to group]. This groups the clients [the selected set of UEs] into separate logical pools called tiers [a plurality of clusters].” Grouping the client into tiers [clusters] requires a policy be determined specifying how to group the clients into tiers.)
and instruct each cluster of the plurality of clusters to perform the FL task. (On page 5 of Chai, “After the selection of clients [each cluster of the plurality of clusters], training proceeds as state-of-the-art FL system does [perform the FL task]”)
Bonawitz and Chai are both related to the same field of endeavor (i.e. federated learning). In view of the teachings of Chai, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to apply the teachings of Chai to Bonawitz in order to improve the speed and accuracy of a federated learning task (see Chai page 12, “Specifically, TIFL achieves an improvement over conventional FL by up to 3× speedup in overall training time and by 6% in accuracy.”
Regarding claim 2 and analogous claim 14:
Bonawitz and Chai teach
The parameter server of claim 1,
Bonawitz further teaches
wherein the scheduler is further configured to: store information on the plurality of subscribing UEs that have each subscribed to perform at least one FL task, (on page 2 of Bonawitz, “Periodically, devices that meet the eligibility criteria (e.g., charging and connected to an unmetered network; see Sec. 3) [the plurality of UEs that have each subscribed to perform at least one FL task] check into the server by opening a bidirectional steam. The stream is used to check liveness and orchestrate multi-step communication [store information].”)
receive, from the at least one module, a request to run or re-run an FL task with respect to a specific ML model, (on page 5 of Bonawitz, “Selectors are responsible for accepting and forwarding device connections. They periodically receive information from the coordinator [the at least one module] about how many devices are needed for each FL population [a request to run or re-run an FL task], which they use to make local decisions about whether or not to accept each device”)
the request specifying at least one condition to be satisfied by UEs performing the requested FL task (on page 5 of Bonawitz, “Selectors are responsible for accepting and forwarding device connections. They periodically receive information from the coordinator about how many devices are needed for each FL population [at least one condition to be satisfied by UEs performing the requested FL task], which they use to make local decisions about whether or not to accept each device”),
determine, using the stored information on the plurality of subscribing UEs, whether the at least one condition is satisfied by the plurality of subscribing UEs, and (on page 5 of Bonawitz, “Selectors are responsible for accepting and forwarding device connections. They periodically receive information from the coordinator about how many devices are needed for each FL population, which they use to make local decisions about whether or not to accept each device [determine whether the condition is satisfied by the plurality of subscribing UEs]”)
transmit, to the at least one module, a response indicating whether the request is granted based on the determination. (on page 5 of Bonawitz, “After the master aggregator and set of Aggregators are spawned, the Coordinator instructs the Selectors to forward a subset of its connected devices [a response indicating whether the request is granted based on the determination] to the Aggregators”)
Bonawitz and Chai are combinable for the same rationale set forth above with respect to claim 1.
Regarding claim 3:
Bonawitz and Chai teach
The parameter server of claim 2,
Bonawitz further teaches:
wherein the at least one condition specifies any one or more of a quality of service (QoS) profile, a minimum number of UEs required to perform the FL task, and a UE hardware capacity requirement for performing the FL task. (on page 5 of Bonawitz, “Selectors are responsible for accepting and forwarding device connections. They periodically receive information from the coordinator about how many devices are needed for each FL population [a minimum number of UEs required to perform the FL task], which they use to make local decisions about whether or not to accept each device”)
Bonawitz and Chai are combinable for the same rationale set forth above with respect to claim 1.
Regarding claim 4:
Bonawitz and Chai teach
The parameter server of claim 3
Bonawitz further teaches
wherein the FL task comprises a plurality of training iterations, and (on page 2 of Bonawitz, “From the potential tens of thousands of devices announcing their availability to the server during a certain time window, the server selects a subset of typically a few hundred who are invited to work on a specific FL task (we discuss the reason for the subsetting in Sec. 2.2). We call this rendezvous between devices and server a round [a training iteration]. Devices stay connected to the server for the duration of the round. The server tells the selected devices what computation to run with an FL plan, a data structure that includes a TensorFlow graph and instructions for how to execute it. Once a round is established, the server next sends to each participant the current global model parameters and any other necessary state as an FL checkpoint (essentially the serialized state of a TensorFlow session). Each participant then performs a local computation based on the global state and its local dataset, and sends an update in the form of an FL checkpoint back to the server. The server incorporates these updates into its global state, and the process repeats. [a plurality of training iterations]")
before a training iteration ends, the scheduler is configured to: check UE status messages received from the UEs in the plurality of clusters to determine whether the UEs are still able to perform the training iteration of the FL task, and (on page 9 of Bonawitz, “We also observe that on average the portion of devices that drop out due to computation errors, network failures, or changes in eligibility varies between 6% and 10%. Therefore, in order to compensate for device drop out as well as to allow stragglers to be discarded, the server typically selects 130% of the target number of devices to initially participate.” Re-grouping the UEs into a plurality of clusters discarding specifically dropped-out devices and stragglers implicitly requires checking status messages to determine whether they have dropped out and making a determination that they are unable to perform the training iteration.)
instruct, responsive to determining at least one UE is unable to perform the training iteration of the FL task, a coordinator to re-group the UEs into a plurality of clusters according to the clustering policy and the UE status messages (on page 9 of Bonawitz, “We also observe that on average the portion of devices that drop out due to computation errors, network failures, or changes in eligibility varies between 6% and 10%. Therefore, in order to compensate for device drop out as well as to allow stragglers to be discarded [re-group the UEs into a plurality of clusters according to the clustering policy and the UE status messages], the server typically selects 130% of the target number of devices to initially participate.” Re-grouping the UEs into a plurality of clusters discarding specifically dropped-out devices and stragglers implicitly requires making a determination that they are unable to perform the training iteration.)
Bonawitz and Chai are combinable for the same rationale set forth above with respect to claim 1.
Claims 5, 6, and 15 are rejected under 35 U.S.C. 103 as being unpatentable over Bonawitz in view of Chai and further in view of Hosseinalipour.
Regarding claim 5:
Bonawitz and Chai teach
The parameter server of claim 4
Bonawitz and Chai fail to teach
wherein the at least one module of the parameter server is configured to: receive, from a UE via the scheduler, a first set of parameters and a second set of parameters, and update the ML model corresponding to the FL task that has been performed by the UEs using the first set and the second set,
wherein the first set of parameters is generated based on a training of a local version of the ML model corresponding to the FL task using at least one training data item stored on the UE,
and wherein the second set of parameters is generated based on a training of a local version of the ML model corresponding to the FL task using at least one training data item generated by the UE and at least one generated coded training data item received from other UEs in the cluster.
Hosseinalipour teaches
wherein the at least one module of the parameter server is configured to: receive, from a UE via the scheduler, a first set of parameters and a second set of parameters, and update the ML model corresponding to the FL task that has been performed by the UEs using the first set and the second set, (on page 5 of Hosseinalipour, “We propose local aggregations at each network layer. To achieve this, each cluster in Fig. 3 follows one of two mechanisms: Distributed Aggregation: The nodes engage in a cooperative scheme facilitated by D2D communications to realize the consensus/average of their local model parameters. The parent node then samples parameters of one of the children [wherein at least one module of the parameter server is configured to receive, from a UE via the scheduler,[…] a second set of parameters] and scales it by the number of children to calculate an approximate sum of the children nodes’ parameters [update the ML model corresponding to the FL task that has been performed by the UEs using […] the second set of parameters.] Instant Aggregation: Each node uploads its local model to the parent node [wherein the at least one module of the parameter server is configured to: receive, from a UE via the scheduler, a first set of parameters]. The parent computes the aggregation directly as a sum off the children nodes’ parameters [update the ML model corresponding to the FL task that has been performed by the UEs using the first set]. See also Fig. 4, and in particular (c) which depicts a network in which the parameter server receive parameters calculated through both Distributed and Instant Aggregation.”)
wherein the first set of parameters is generated based on a training of a local version of the ML model corresponding to the FL task using at least one training data item stored on the UE, (on page 1 of Hosseinalipour, “Each round of model training consists of two steps: (i) local updating [training of a local version of the ML model corresponding to the FL task], where each device updates its local model [the first set of parameters] based on its dataset [at least one training data item stored on the UE] and the global model, e.g., using gradient descent, and (ii) global aggregation, where the server gathers devices’ local models and computes a new global model, which is then synchronized across the devices to begin the next round”)
wherein the second set of parameters is generated based on a training of a local version of the ML model corresponding to the FL task using at least one training data item generated by the UE and at least one generated training data item received from other UEs in the cluster (on page 5 of Hosseinalipour, “We propose local aggregations at each network layer. To achieve this, each cluster in Fig. 3 follows one of two mechanisms: Distributed Aggregation: The nodes engage in a cooperative scheme facilitated by D2D communications [generated based on a training of a local version of the ML model corresponding to each FL task using at least one training data item generated by the UE and at least one generated training data item received from other UEs in the cluster] to realize the consensus/average [the second set of parameters] of their local model parameters . The parent node then samples parameters of one of the children and scales it by the number of children to calculate an approximate sum of the children nodes’ parameters. Instant Aggregation: Each node uploads its local model to the parent node. The parent computes the aggregation directly as a sum off the children nodes’ parameters.”)
Bonawitz and Chai are combinable for the same rationale set forth above with respect to claim 1.
Bonawitz, Chai, and Hosseinalipour are each related to the same field of endeavor (i.e. performing perception tasks using ensembles of neural networks. In view of the teachings of Hosseinalipour, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to apply the teachings of Hosseinalipour to Bonawitz and Chai in order to perform federated learning in a more energy-efficient manner (On page 14 of Hosseinalipour, “The accumulated energy consumption of the edge devices through the training phase is depicted in Fig. 14, which reveals around 50% energy saving on average as compared to the EUT baseline.”).
Regarding claim 6:
Bonawitz, Chai, and Hosseinalipour teach
The parameter server of claim 5
Bonawitz further teaches
wherein when the first set of parameters has been received from a pre-defined minimum number of UEs from the plurality of clusters within a pre-defined time period, the at least one module is configured to update the ML model by: aggregating the first set of parameters received from the UEs in the plurality of clusters; and updating the ML model using the aggregated first set of parameters. (on page 3 of Bonawitz, “The server waits for the participating devices to report updates [the first set of parameters]. As updates are received, the server aggregates them using Federated Averaging [aggregating the first set of parameters received from the UEs in the plurality of clusters;] and instructs the reporting devices when to reconnect (see also Sec. 2.3). If enough [a pre-defined number of UEs] devices report in time [within a pre-defined time period], the round will be successfully completed and the server will update its global model [and updating the ML model using the aggregated first set of parameters], otherwise the round is abandoned”)
Bonawitz, Chai and Hosseinalipour are combinable for the same rationales set forth above with respect to claim 5.
Regarding claim 15:
Bonawitz and Chai teach
The method of claim 14
Bonawitz and Chai fail to teach
Further comprising: transmitting, to a coordinator for coordinating the execution of FL tasks by the UEs in each cluster a request to determine a per-cluster data coding optimization policy to be used by each UE in the plurality of clusters
wherein the per-cluster data coding optimization policy defines how UEs within each cluster transmit data
Husseinalipour teaches
Further comprising: transmitting, to a coordinator for coordinating the execution of FL tasks by the UEs in each cluster a request to determine a per-cluster data coding optimization policy to be used by each UE in the plurality of clusters (on page 11 of Hosseinalipour, see Algorithm 2:
PNG
media_image4.png
744
1111
media_image4.png
Greyscale
)
wherein the per-cluster data coding optimization policy defines how UEs within each cluster transmit data (on page 11 of Hosseinalipour, see Algorithm 2 above).
Bonawitz, Chai and Hosseinalipour are combinable for the same rationales set forth above with respect to claim 5.
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over Bonawitz in view of Chai and Gupta (EP 3869847 A1). Bonawitz and Chai teach The parameter server of claim 1. wherein Bonawitz and Chai fail to teach the cellular network is an open radio-acess network (ORAN), the parameter server is a service management and orchestration (SMO) platform comprising a non-real time radio intelligent controller (non-RT-RIC), the at least one module for updating the at least one ML model is a software application (rAPP) configured to run on the non-RT-RIC, and the scheduler is a software application configured to run on the non-RT-RIC. Gupta teaches the cellular network is an open radio-access network (ORAN), (in paragraph [0090] of Gupta, “FIG. provides a high-level view of an Open RAN (O-RAN) architecture 500. The O-Ran architecture 500 [the cellular network is an open radio-access network (ORAN)] includes four O-RAN defined interfaces – namely, the A1 interface, the O1 interface, the O2 interface, and the Open Fronthall Management (M)-plane interface – which connect the Service Management and Orchestration (SMO) framework 502 to O-RAN network functions (NFs) 503 and the O-Cloud 506.”) the parameter server is a service management and orchestration (SMO) platform comprising a non-real time radio intelligent controller (non-RT-RIC), (in paragraph [0105] of Gupta, “In some embodiments, the Non-RT RIC 612 is a function that sits within [comprising a non-real time radio intelligent controller] the SMO platform (or SMO framework) 602 [a service management and orchestration platform] in the O-RAN architecture. The primary goal of non-RT RIC is to support intelligent radio resource management for a non-real-time interval (i.e., greater than 500 ms), policy optimization in RAN, and insertion of AI/ML models to near-RT RIC and other RAN functions.”) the at least one module for updating the at least one ML model is a software application (rAPP) configured to run on the non-RT-RIC, and the scheduler is a software application configured to run on the non-RT-RIC (in paragraph [0107] of Gupta, “The non-RT RIC 612 can be an ML training host to host the training of one or more ML models. ML training can be performed offline using data collected from the RIC, O-DU 615, and O-RU 616. For supervised learning, non-RT RIC 612 is part of the SMO 602, and the ML training host [the at least one module for updating the at least one ML model] and/or ML model host/actor can be part of the non-RT RIC 612 [configured to run on the non-RT-RIC] and/or the near-RT RIC 614. In some implementations, the non-RT RIC 612 [the scheduler] may request or trigger ML model training in the training hosts regardless of where the model is deployed and executed” Bonawitz, Chai, and Hosseinalipour are combinable for the same rationales set forth above with respect to claim 5. Bonawitz, Chai, Hosseinalipour, and Gupta are all related to the same field of endeavor (i.e., performing machine learning tasks over cellular networks). In view of the teachings of Gupta it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to apply the teachings of Gupta to Bonawitz and Chai in order to practically implement their methods for federated learning in a ORAN architecture (in paragraph [0005] of Gupta, “Such enhanced operations can include techniques for multi-access management in O-RAN architectures”).
Claim 11 is rejected under 35 U.S.C. 103 as being unpatentable over Hosseinalipour in view of Chai.
Hosseinalipour teaches
The UE of claim 10
Hosseinalipour further teaches
generate a second set of parameters, by training a local version of the ML model corresponding to the FL task using at least one generated coded training data item generated by the UE and at least one generated coded training data item received from other UEs in the cluster of UEs, and, (on pages 6 and 7 of Hosseinalipour, “In LUT clusters [the cluster of UEs], devices leverage D2D communications to obtain an approximate value of the average of their parameters. A basic approach would be to implement a message passing algorithm where nodes exchange parameters [at least one generated coded training data item received from other UEs in the cluster of UEs], with their neighbors until each node in the cluster has all parameters [using at least one generated coded training data item generated by the UE and at least one generated coded training data item received from other UEs in the cluster of UEs; each node has both its own parameters and the parameters of all other (at least one) nodes in the cluster] stored locally. Each node can then readily calculate [training a local version of the ML model corresponding to the FL task] the aggregated value [a second set of parameters] and one of them can be sampled by the parent node.
transmit […] the second set of parameters to the parameter server (on page 5 of Hosseinalipour, “The parent node [the parameter server] then samples [transmit] parameters [the second set of parameters] of one of the children and scales it by the number of children to calculate an approximate sum of the children nodes’ parameters”)
Hoseeinalipour fails to teach
where the at least one processor is further configured to: generate a first set of parameters, by training a local version of the ML model corresponding to the FL task using at least one training data item stored on the UE,
Chai teaches
where the at least one processor is further configured to: generate a first set of parameters, by training a local version of the ML model corresponding to the FL task using at least one training data item stored on the UE, (on page 7 of Chai, "In each training round, 5 clients are selected to train [training a local version of the ML model corresponding to the FL task] on their own data [using at least one training data item stored on the UE] and send the trained weights to the server which aggregates them and updates the global model")
transmit the first set of parameters […] to the parameter server (on page 7 of Chai, "In each training round, 5 clients are selected to train on their own data and send the trained weights to the server [transmit the first set of parameters [..] to the parameter server] which aggregates them and updates the global model")
Hosseinalipour and Chai are both related to the same field of endeavor (i.e. federated learning). In view of the teachings of Chai, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to apply the teachings of Chai to Hosseinalipour in order to improve the speed and accuracy of a federated learning task (see Chai page 12, “Specifically, TIFL achieves an improvement over conventional FL by up to 3× speedup in overall training time and by 6% in accuracy.”
Claim 12 is rejected under 35 U.S.C. 103 as being unpatentable over Hosseinalipour in view of Bonawitz and Chai.
Hosseinalipour and Chai teach
The UE of claim 11
Hosseinalipour and Chai fail to teach
wherein the at least one processor is further configured to: transmit, to the parameter server, a subscription request indicating the UE is able to perform at least one FL task; and
periodically transmit, to the parameter server, status update messages.
Bonawitz teaches
wherein the at least one processor is further configured to: transmit, to the parameter server, a subscription request indicating the UE is able to perform at least one FL task; and (on page 4 of Bonawitz, “Upon invocation by the job scheduler in a separate process, the FL runtime contacts the FL server [transmit, to the parameter server] to announce that it is ready to run tasks for the given FL population [a subscription request indicating the UE is able to perform at least one FL task]”)
periodically transmit, to the parameter server, status update messages. (on page 4 of Bonawitz, “This schedules a periodic FL runtime job using Android’s JobScheduler… After FL plan execution [periodically], the FL runtime reports computer updates and metrics to the server [transmit, to the parameter server, status update messages] and cleans up any temporary resources”)
Hosseinalipour and Chai are combinable for the same rationale set forth above with respect to claim 11.
Hosseinalipour, Chai, and Bonawitz are each related to the same field of endeavor (i.e. federated learning). In view of the teachings of Bonawitz, it would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to apply the teachings of Bonawitz to Hosseinalipour and Chai in order to implement the ideas of Hosseinalipour and Chai in a large-scale federated learning application (on page 9 of Bonawitz, “We designed the FL system to elastically scale with the number and sizes of the FL populations, potentially up into the billions. Currently the system is handling a cumulative FL population size of approximately 10M daily active devices, spanning several different applications.”)
Claims 16 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Chai in view of Bonawitz.
Regarding claim 16:
Chai teaches
A method performed by a coordinator for training machine learning (ML) models in a cellular network, the method comprising: receiving, from a parameter server, (on page 4 of Chai, “The overall system architecture of TIFL is present in Fig. 2. TIFL follows the system design to the state-of-the-art FL system and adds two new components: a tiering module (a profile and tiering algorithms) and a tier scheduler. These newly added components can be incorporated into the coordinator [performed by a coordinator for training machine learning (ML models in a cellular network].”)
receiving, from a parameter server, a clustering policy for a federated learning (FL) task and information on a set of user equipments (UEs), and (on page 4 of Chai, ”In TIFL, the first step is to collect the latency metrics of all the available clients [information on a set of user equipments (UEs)] through a lightweight profiling as detailed in section 4.2. The profiled data is further utilized by our tiering algorithm [a clustering policy for a federated learning (FL) task]. This groups the clients into separate logical pools called tiers.” Grouping the client into tiers [clusters] requires receiving from a parameter server a clustering policy specifying how to group the clients into tiers)
grouping the set of UEs into a plurality of clusters based on the clustering policy, and wherein the clustering policy specifies how to group the set of UEs into the plurality of clusters, and (on page 4 of Chai, ”In TIFL, the first step is to collect the latency metrics of all the available clients through a lightweight profiling as detailed in section 4.2. The profiled data is further utilized by our tiering algorithm [the clustering policy]. This groups the clients [the set of UEs] into separate logical pools called tiers [a plurality of clusters].”)
Chai fails to teach
wherein the set of UEs is selected from a plurality of UEs in the cellular network as being suitable for performing the FL task
Bonawitz teaches
wherein the set of UEs is selected from a plurality of UEs in the cellular network as being suitable for performing the FL task (on page 2 of Bonawitz, “The server selects [selected from a plurality of UEs in the cellular network] a subset of connected devices [the set of UEs] based on certain goals like the optimal number of participating devices [as being suitable for performing the FL task] (typically few hundred devices participate in each round”.)
Chai and Bonawitz are combinable for the same rationale set forth above with respect to claim for the same reasons as set out above with respect to claim 12.
Regarding claim 19:
Chai and Bonawitz teach
The method of claim 16
Chai further teaches
further comprising: transmitting, to the parameter server, information on which UEs are in each cluster for the FL task, so that the parameter server is able to instruct the UEs in each cluster to perform the FL task (See Figure 2 in Bonawitz:
PNG
media_image5.png
465
565
media_image5.png
Greyscale
The Tier Scheduler [coordinator] transmits cluster groups to the Aggregator [part of the parameter server].)
Chai and Bonawitz are combinable for the same rationale set forth above with respect to claim for the same reasons as set out above with respect to claim 12.
Claims 17 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Chai in view of Bonawitz and Hosseinalipour.
Regarding claim 17:
Chai and Bonawitz teach The method of claim 16
Chai and Bonawitz fail to teach
further comprising, determining a per-cluster data coding optimization policy to be used by each UE in the plurality of clusters when performing the Fl task,
wherein the per-cluster data coding optimization policy defines how UEs within each cluster transmit data
Hosseinalipour teaches
further comprising, determining a per-cluster data coding optimization policy to be used by each UE in the plurality of clusters when performing the FL task, (on page 11 of Hosseinalipour, see Algorithm 2.
PNG
media_image6.png
744
1044
media_image6.png
Greyscale
)
wherein the per-cluster data coding optimization policy defines how UEs within each cluster transmit data (on page 11 of Hosseinalipour, Algorithm 2 (see above). )
Chai, Bonawitz, and Hosseinalipour are combinable for the same rationale set forth above with respect to claim for the same reasons as set out above with respect to claim 12.
Regarding claim 18:
Chai, Bonawitz, and Hosseinalipour teach
The method of claim 17
Chai further teaches
further comprising: periodically re-grouping, after a pre-defined time period, the set of UEs into a plurality of clusters based on the clustering policy, while the FL task is being executed; and (on page 4 of Chai, ”In TIFL, the first step is to collect the latency metrics of all the available clients through a lightweight profiling as detailed in section 4.2. The profiled data is further utilized by our tiering algorithm [the clustering policy]. This groups the clients [the set of UEs] into separate logical pools called tiers [a plurality of clusters].”)
Hosseinalipour further teaches
re-determining, after the re-grouping, a per-cluster data coding optimization policy to be used by each UE in the plurality of clusters when performing the FL task. (on page 11 of Hosseinalipour, see Algorithm 2:
PNG
media_image7.png
744
1141
media_image7.png
Greyscale
)
Chai, Bonawitz, and Hosseinalipour are combinable for the same rationale set forth above with respect to claim for the same reasons as set out above with respect to claim 12.
Claim 20 is rejected under 35 U.S.C. 103 as being unpatentable over Chai in view of Bonawitz and Gupta.
Chai and Bonawitz teach
The method of claim 16
Chai and Bonawitz fail to teach
wherein the coordinator is a software application configured to run on a near-real time radio intelligent controller (near-RT-RIC)
which controls nodes of the cellular network
Gupta teaches
wherein the coordinator is a software application configured to run on a near-real time radio intelligent controller (near-RT-RIC) which controls nodes of the cellular network
(in paragraph [0107] of Gupta, “The non-RT RIC 612 can be an ML training host to host the training of one or more ML models. ML training can be performed offline using data collected from the RIC, O-DU 615, and O-RU 616. For supervised learning, non-RT RIC 612 is part of the SMO 602, and the ML training host and/or ML model host/actor [a software application] can be part of the non-RT RIC 612 and/or the near-RT RIC 614 [configured to run on a near-real time intelligent controller (near-RT-RIC)]. In some implementations, the non-RT RIC 612 may request or trigger ML model training in the training hosts regardless of where the model is deployed and executed” The broadest reasonable interpretation of controls nodes in the cellular network includes a ML training host running on the near-RT-RIC; the training host is a node in the cellular network).
Chai and Bonawitz are combinable for the same rationale set forth above with respect to claim for the same reasons as set out above with respect to claim 12.
Chai, Bonawitz, and Gupta are combinable for the same rationale set forth above with respect to claim 9.
Conclusion
Any inquiry concerning this communication or earlier communication from the examiner should be directed to Alma Thompson whose telephone number is +1 (571) 270-1810. The examiner can normally be reached Monday-Friday, 9:00 am – 5:00 pm EST. If attempts to reach the examiner are unsuccessful, the examiner’s supervisor, Michael Huntley can be reached at +1 (303) 297-4307.
Should applicant desire to communicate via email or schedule an Applicant-initiated interview, the Applicant may use the Automated Interview Request (AIR) form online at <http://www.uspto.gov/patent/uspto-automated-interview-request-air-form> In particular, the AIR form allows the Applicant to include authorization to communicate via email with the examiner for interview-related communications.
Information regarding the status of an application may be obtained from the Patent Application 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://pairdirect.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.
/AT/Examiner, Art Unit 2129
/MICHAEL J HUNTLEY/Supervisory Patent Examiner, Art Unit 2129
Prosecution Insights