DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Amendments
This Office Action is in response to the amendment filed on January 30, 2026.
Claims 1, 10-17, and 19 have been amended.
No claims have been cancelled.
No new claims have been added.
The objections and rejections from the prior correspondence that are not restated herein are withdrawn.
Response to Arguments
Applicant's arguments filed on January 30, 2026 have been fully considered.
Applicant’s arguments regarding 35 USC 112(a) rejections have been fully considered, but are not persuasive. Applicant argues:
“Claim 10 and its dependent claims are amended as necessary to recite a processor and memory which generate software modules as indicated in the claims. The Applicant urges that limitations in this form are enabled by the present specification.”
Examiner respectfully disagrees. While the amendments now recite sufficient structure for performing the claimed function, the amendments introduce a new 112(a) rejection regarding the generation step of the units. The specification lacks written description for how the units recited in claims 10-17 are generated, as shown in the 112(a) rejection below.
Applicant’s arguments regarding 35 USC 101 rejections have been fully considered, but are not persuasive. Applicant argues:
“The Applicant respectfully submits that: (1) the claims are not directed to an abstract idea; and (2) in the alternative, the claims integrate a recited judicial exception into a practical application. See 2019 Revised Patent Subject Matter Eligibility Guidance, Vol. 84, No. 4 Federal Register, p. 50 (January 7, 2019). As such, the Applicant respectfully asserts that claims 1 and 12 are directed to statutory subject matter, and consequently claims 1-19 are patent-eligible.”
Examiner respectfully disagrees. Claims 1 and 10 recite the following judicial exceptions:
determining/determine, for a layer of the multi-layer NN model based on the codeword, that a weight matrix of the neural network model is the first weight matrix, […] (Mental process – A person can mentally determine that a weight matrix is the first weight matrix created – see MPEP § 2106.04(a)(2)(III))
updating/update the codeword when a preset stop condition is not met, to obtain an updated codeword; (Mental process – see MPEP § 2106.04(a)(2)(III))
determining/determining, by using the updated codeword obtained from the memory, to create a second weight matrix, (Mental process – A person can mentally determine (e.g., decide) to create a weight matrix by observing and evaluating a codeword – see MPEP § 2106.04(a)(2)(III))
Applicant further argues:
“The Applicant respectfully submits that the present claims do not recite mathematical concepts, methods of organizing human activities, or mental processes as described above. And there is no evidence that approval has been granted by the Technology Center Director even though the claims do not actually reflect mental processes. While the Action alleges that the steps performed by the computer are mental processes, the Applicant urges that the specific steps comprise operations that only a computer can perform including retrieving data from memory registers and performing computer-based comparisons, digit by digit or word by word, to reach a conclusion as required by the various elements of the claim. Many such operations are even performed in hardware rather than software. For one example of a claim element that cannot be created by the human mind, claim 1 as presently constituted requires "obtaining a codeword from a memory to create a first weight matrix of a neural network model". Therefore, the claims do not recite a judicial exception, are not directed to a judicial exception, and the eligibility analysis ends. Thus, the Applicant respectfully requests that the patent-eligibility rejection be withdrawn.”
Examiner respectfully disagrees. The examiner identified judicial exceptions recited in the claims that are reasonably within the capabilities of the human mind, such as determining that a weight matrix is the first weight matrix, updating a codeword, and determining (e.g., deciding) to create a second weight matrix. The additional elements recited in the claim, such as obtaining a codeword from memory, for example, is considered insignificant extra solution activity of mere data gathering, as shown in the 101 rejections below. The remaining additional limitations of claims 1 and 10 are considered mere instructions to apply the judicial exception using generic computer components. For this reason, claims 1 and 10 are directed to an abstract idea.
Applicant further argues:
“The claim requires, as noted above, the use of codewords to create a weight matrix which is then trained (modified) with training data. As such, this removes the need to transmit one or more weight matrices for each layer of a multi-layer NN model. As stated in para. 0006 of the specification, this process
... reduce[s] a data amount of a weight matrix existing in a process of training a neural network model and a calculation amount of an intermediate parameter existing in a process of updating the weight matrix, so that when a preset task (for example, image classification or text translation) is executed by using the neural network model, a memory bottleneck problem can be resolved, and an expected effect can be achieved.
Furthermore, as stated in para. 0008,
In comparison with a conventional technology, in this embodiment of this disclosure, when the neural network model is trained, the weight matrix is no longer directly read from the memory, but the codeword corresponding to the weight matrix is read, to form a weight matrix for training. Memory space occupied by the codeword is far less than memory space occupied by the weight matrix, and therefore a data amount read from the memory can be greatly reduced, and a memory bottleneck problem can be overcome. In addition, in this disclosure, in a model training process, an update amount of the weight matrix is no longer calculated, but an update amount of the codeword is calculated, to determine a new weight matrix for subsequent training. Therefore, a calculation amount of an intermediate parameter existing in an updating process can be reduced, so that the neural network model can be smoothly trained in a resource-limited scenario.
Again, as suggested above, these savings are realized for each layer of a NN model which can easily be 15 layers or (much) more. Thus, there is a computer specific benefit to the inventive subject matter.
Claim 1 is directed to a neural network model training method. Claims 10 and 19 are directed to neural network model training apparatus/device but have similar limitations. As shown above in relation to claim 1 and to the cited text from the present Specification, the invention is directed to a practical application because memory consumption and data transmission bottlenecks can be reduced as a part of training a NN model. Thus, while the Applicant strenuously argues that this is nothing but a computer centric problem relating to computer specific data storage and transmission problems/inefficiencies and is not anything a human could or would do, the fact is that the solution is a practical application in the computer world. The rejection under Section 101 should be removed because the claims are clearly subject matter eligible because they are directed to a practical application by virtue of the computer specific efficiencies as noted above.”
Examiner respectfully disagrees. It should be noted that claim 1 limitation of obtaining a codeword from a memory to create a first weight matrix of a neural network model is not considered as an abstract idea, but rather insignificant extra solution of data gathering under Step 2 Prong 2. Under Step 2B, this limitation is considered well-understood, routine, and conventional activity (see MPEP § 2106.05(d)(ll)(i) - Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information)). Additionally, applicant’s alleged improvement is merely the expected result of using a smaller representation of the weight matrix (i.e., codeword). Therefore, reading a codeword instead of a weight matrix from memory reduces data size, but does not constitute an improvement to computer functionality or amount to significantly more than the judicial exception.
Applicant’s arguments regarding 35 USC 103 rejections have been fully considered but are moot because the new grounds of rejection teach the amended limitations. It should be noted that LIU in view of LI teaches the newly added limitations of independent claims 1, 10, and 19 as shown in the 103 rejections below.
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 10-17 are 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.
Regarding Claim 10:
Claim 10 recites generating a first obtaining unit, generating a first training unit, generating a first updating unit, generating a storage unit, generating a second training unit, and generating a stop unit. The specification does not describe how any these units recited in claim 10 are generated, and thus the specification lacks of written description for the generation of these units. For purposes of examination, generates will be construed as includes, which is supported by the specification (see paragraphs [00192-00225]).
Regarding Claims 11-17:
Claim 11 recites generates: a division unit.
Claim 12 recites generates: a first division subunit; a clustering subunit; a first determining subunit.
Claim 13 recites generates: a dimension reduction subunit; a second grouping subunit; a calculation subunit.
Claim 14 recites generates: a releasing unit.
Claim 15 recites generates: a second determining subunit; a third determining subunit.
Claim 16 recites generates: a first obtaining subunit; a second obtaining subunit; a third obtaining subunit.
Claim 17 recites generates: a second obtaining subunit.
The specification does not describe how any these units recited in claims 11-17 are generated, and thus the specification lacks of written description for the generation of these units. For purposes of examination, generates will be construed as includes, which is supported by the specification (see paragraphs [00192-00225]).
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.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 10-18 are 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.
Regarding Claim 10:
The claim recites “A neural network model training apparatus, wherein the device comprises a memory and a processor […].” However, the device lacks antecedent basis and it is not clear if the apparatus includes a device comprising a memory and a processor, or if the apparatus and device are separate elements. For purposes of examination, the apparatus will be treated as including a device comprising a memory and a processor.
Regarding Claims 11-18:
The dependent claims inherit the deficiencies of their respective parent claims and are likewise rejected.
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 1-19 are rejected under 35 U.S.C.101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1: Claims 1-9 are directed to a process. Claims 10-19 are directed to a machine or an article of manufacture.
With respect to claim(s) 1, 10, and 19:
2A Prong 1: The claim(s) recite(s) an abstract idea. Specifically:
determining/determine, for a layer of the multi-layer NN model based on the codeword, that a weight matrix of the neural network model is the first weight matrix, […] (Mental process – A person can mentally determine that a weight matrix is the first weight matrix created – see MPEP § 2106.04(a)(2)(III))
updating/update the codeword when a preset stop condition is not met, to obtain an updated codeword; (Mental process – see MPEP § 2106.04(a)(2)(III))
determining/determining, by using the updated codeword obtained from the memory, to create a second weight matrix, (Mental process – A person can mentally determine (e.g., decide) to create a weight matrix by observing and evaluating a codeword – see MPEP § 2106.04(a)(2)(III))
If claim limitations, under their broadest reasonable interpretation, cover performance of the limitations as a mental process, but for the recitation of generic computer components, then the claim limitations fall within the mathematical or mental process grouping of abstract ideas. Accordingly, the claim “recites” an abstract idea.
2A Prong 2: The additional elements recited in the claim(s) do not integrate the abstract idea into a practical application, individually or in combination.
Additional elements:
(Claim 1) An efficient neural network (NN) model training method for training a multi-layer NN model, wherein the method comprises: (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
(Claim 10) A neural network model training apparatus, wherein the device comprises a memory and a processor; (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
(Claim 10 and 19) the memory is configured to store instructions; and the processor is configured to execute the instructions in the memory, to perform the neural network model training method, wherein the method comprises: (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
(Claim 19) wherein the device comprises a memory and a processor (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
(Claim 10) a first obtaining unit, a first training unit, an updating unit, a storage unit, a second training unit, and a stop unit. (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
obtaining/obtain a codeword from a memory to create a first weight matrix of a neural network model; (Mere data gathering – Adding insignificant extra-solution activity of mere data gathering to the judicial exception – see § MPEP2106.05(g).)
[…] and training the first weight matrix for the layer by using training data; (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
storing/store the updated codeword in the memory; (Adding insignificant extra-solution activity to the judicial exception – see § MPEP2106.05(g).)
[…] and training the second weight matrix by using the training data; (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
stopping/stop training of the neural network model when the preset stop condition is met. (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Starting/stopping training according to a preset stop condition describes a generic training process of a neural network – see MPEP 2106.05(f).)
Since the claim as a whole, looking at the additional elements individually and in combination, does not contain any other additional elements that are indicative of integration into a practical application, the claim is directed to an abstract idea.
2B: The claim(s) do(es) not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
(Claim 1) An efficient neural network (NN) model training method for training a multi-layer NN model, wherein the method comprises: (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
(Claim 10) A neural network model training apparatus, wherein the device comprises a memory and a processor; (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
(Claim 10 and 19) the memory is configured to store instructions; and the processor is configured to execute the instructions in the memory, to perform the neural network model training method, wherein the method comprises: (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
(Claim 19) wherein the device comprises a memory and a processor (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
(Claim 10) a first obtaining unit, a first training unit, an updating unit, a storage unit, a second training unit, and a stop unit. (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
obtaining/obtain a codeword from a memory to create a first weight matrix of a neural network model; (Data gathering is well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception (WURC)- see MPEP § 2106.05(d)(ll)(i) - Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information).)
[…] and training the first weight matrix for the layer by using training data; (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
storing/store the updated codeword in the memory; (Storing and retrieving information in memory is well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception (WURC)- see MPEP § 2106.05(d)(ll)(iv) - Storing and retrieving information in memory, Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015); OIP Techs., 788 F.3d at 1363, 115 USPQ2d at 1092-93.)
[…] and training the second weight matrix by using the training data; (Mere instructions to apply an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
stopping/stop training of the neural network model when the preset stop condition is met. (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. Starting/stopping training according to a preset stop condition describes a generic training process of a neural network – see MPEP 2106.05(f).)
Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
With respect to claim(s) 2 and 11:
2A Prong 1: The claims recite an abstract idea. Specifically:
dividing/divide initial weight matrix, to determine a codeword corresponding to the initial weight matrix. (Mathematical calculations and/or mental process – see MPEP § 2106.04(a)(2))
2A Prong 2: The additional elements recited in the claims do not integrate the abstract idea into a practical application, individually or in combination.
Additional elements:
(Claim 11) a division unit […]; (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
2B: The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
(Claim 11) a division unit […]; (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
Therefore, the claims are ineligible.
With respect to claim(s) 3 and 12:
2A Prong 1: The claim recites an abstract idea. Specifically:
dividing/divide the initial weight matrix into k submatrices of a same dimension, wherein k is a positive integer greater than 1; (Mathematical calculations and/or mental process – see MPEP § 2106.04(a)(2))
performing/perform clustering processing on the k submatrices of a same dimension, to obtain n codewords corresponding to the k submatrices of a same dimension, wherein n is a positive integer greater than 0, and n ≤ k; (Mathematical calculations – see MPEP § 2106.04(a)(2)(I))
determining/determine the n codewords as codewords corresponding to the initial weight matrix. (Mental process – see MPEP § 2106.04(a)(2)(III))
2A Prong 2: The additional elements recited in the claim do not integrate the abstract idea into a practical application, individually or in combination.
Additional elements:
(Claim 12) a first division subunit, a clustering subunit, and a first determining subunit. (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
(Claim 12) a first division subunit, a clustering subunit, and a first determining subunit. (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
Therefore, the claims are ineligible.
With respect to claim(s) 4 and 13:
2A Prong 1: The claims recite an abstract idea. Specifically:
reducing/reduce each of the k submatrices of a same dimension into a one-dimensional vector, to obtain k one-dimensional vectors; (Mathematical calculations – see MPEP § 2106.04(a)(2)(I))
grouping/group the k one-dimensional vectors into n vector groups, wherein each vector group comprises at least one one-dimensional vector; (Mathematical calculations and/or mental process – see MPEP § 2106.04(a)(2))
performing/perform average calculation on element values at corresponding locations in all one-dimensional vectors that belong to an ith vector group in the k one-dimensional vectors, to obtain a codeword corresponding to all the one-dimensional vectors in the ith vector group, wherein i is an integer that ranges from 1 to n. (Mathematical calculations – see MPEP § 2106.04(a)(2)(I))
2A Prong 2: The additional elements recited in the claims do not integrate the abstract idea into a practical application, individually or in combination.
Additional elements:
(Claim 13) a dimension reduction subunit, a second grouping unit, and a calculation subunit (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
2B: The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
(Claim 13) a dimension reduction subunit, a second grouping unit, and a calculation subunit (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
Therefore, the claims are ineligible.
With respect to claim(s) 5 and 14:
2A Prong 2: The additional elements recited in the claims do not integrate the abstract idea into a practical application, individually or in combination.
Additional elements:
(Claim 14) a releasing unit (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
releasing/release the weight matrix of the neural network model in the memory when the preset stop condition is not met. (Adding insignificant extra-solution activity to the judicial exception – see § MPEP2106.05(g).)
2B: The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
(Claim 14) a releasing unit (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
releasing/release the weight matrix of the neural network model in the memory when the preset stop condition is not met. (Releasing memory at the end of a computation is incidental to the primary process and also does not provide an inventive concept, particularly when the activity is a well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception (WURC)- see MPEP § 2106.05(d)(ll)(i) - Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information).)
Therefore, the claims are ineligible.
With respect to claim(s) 6 and 15:
2A Prong 1: The claims recite an abstract idea. Specifically:
determining/determine a weight gradient of the first weight matrix of the neural network model when the preset stop condition is not met; (Mathematical calculations – see MPEP § 2106.04(a)(2)(I))
determining/determine a codeword gradient based on the first weight gradient, and determining the updated codeword based on the codeword gradient. (Mathematical calculations – see MPEP § 2106.04(a)(2)(I))
2A Prong 2: The additional elements recited in the claims do not integrate the abstract idea into a practical application, individually or in combination.
Additional elements:
(Claim 15) a second determining subunit and a third determining subunit (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
2B: The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
(Claim 15) a second determining subunit and a third determining (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
Therefore, the claims are ineligible.
With respect to claim(s) 7 and 16:
2A Prong 1: The claims recite an abstract idea. Specifically:
performing/perform weighted summation on weight gradients that are in the weight gradient and that are of submatrices corresponding to index numbers that belong to a jth codeword, to obtain a codeword gradient corresponding to the jth codeword, wherein j is an integer that ranges from 1 to n; (Mathematical calculations – see MPEP § 2106.04(a)(2)(I))
optimizing/optimize the codeword gradient corresponding to the jth codeword, to obtain an update amount of the jth codeword; (Mathematical calculations – see MPEP § 2106.04(a)(2)(I))
updating/update the jth codeword by using the update amount of the jth codeword, to obtain an updated jth codeword. (Mathematical calculations – see MPEP § 2106.04(a)(2)(I))
2A Prong 2: The additional elements recited in the claims do not integrate the abstract idea into a practical application, individually or in combination.
Additional elements:
(Claim 16) a first obtaining subunit¸ a second obtaining subunit, and a third obtaining subunit (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
2B: The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
(Claim 16) a first obtaining subunit¸ a second obtaining subunit, and a third obtaining subunit (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
Therefore, the claims are ineligible.
With respect to claim(s) 8 and 17:
2A Prong 2: The additional elements recited in the claims do not integrate the abstract idea into a practical application, individually or in combination.
Additional elements:
obtaining an index, wherein the index is a correspondence between the codeword and the weight matrix of the neural network model (Mere data gathering – see § MPEP2106.05(g).)
(Claim 17) a second obtaining subunit (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
2B: The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
obtaining an index, wherein the index is a correspondence between the codeword and the weight matrix of the neural network model (Simply appending well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception (WURC)- see MPEP § 2106.05(d)(ll)(i) - Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information).)
(Claim 17) a second obtaining subunit (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea – see MPEP 2106.05(f).)
Therefore, the claims are ineligible.
With respect to claim(s) 9 and 18:
2A Prong 1: The claims recite an abstract idea. Specifically:
a difference between a result label value corresponding to the training data and a result output by the neural network model for the training data is less than a preset difference; (Mathematical calculations – see MPEP § 2106.04(a)(2)(I))
a change rate of the difference between the result label value corresponding to the training data and the result output by the neural network model for the training data is less than a preset change threshold; (Mathematical calculations – see MPEP § 2106.04(a)(2)(I))
a quantity of update times of a model parameter in the neural network model reaches a preset quantity of update times; (Mathematical calculations and/or mental process – see MPEP § 2106.04(a)(2))
an output value of a loss function used by the neural network model reaches a preset threshold, wherein the loss function is used to measure the difference between the result output by the neural network model for the training data and the result label value corresponding to the training data. (Mathematical calculations – see MPEP § 2106.04(a)(2)(I))
Additionally, the claims do not recite any new additional elements that would amount to an integration of the abstract idea into a practical application (individually or in combination) or significantly more than the judicial exception.
Therefore, the claims are ineligible.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-2, 8, 10-11, 17, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over LIU (EP 3627397 A1) in view of LI (US 20160307095 A1), hereafter LIU and LI respectively.
Regarding Claim 1:
LIU teaches:
An efficient neural network (NN) model training method for training a multi-layer NN model, wherein the method comprises: (LIU [ 0271] teaches: "a training algorithm of the multilayer artificial neural network computation are solved.")
updating the codeword when a preset stop condition is not met, to obtain an updated codeword; (LIU [0190] teaches training the weight codebook containing the central weights using a back propagation algorithm (i.e., updating the codeword […] to obtain an updated codeword). LIU [0611] teaches training using back propagation is repeated until a loss function converges to a first threshold value or until the count of training times is more than or equal to a second threshold value (i.e., when a preset stop condition is not met), and updating the target model according to the loss function. The target model is the compression neural network model that has been compressed using the weight codebook method.)
storing the updated codeword in the memory; (LIU [0645] teaches a memory storing the a compression neural network model, which has been compressed by implementing the weight codebook method. This memory is used during the training to store the model and perform back propagation.)
determining, by using the updated codeword obtained from the memory to create a second weight matrix, and training the second weight matrix by using training data; and (LIU [0005], [0190], [0611], and [0645] teaches repeating the back propagation process to updated the model until a condition is met. This repeating process updates the weight codebook containing the central weights, and with each count of training time and/or updating the model according to the loss function, results in new weights for the weight codebook, and thus creating a second weight matrix.)
stopping training of the neural network model when the preset stop condition is met. (LIU [0611] teaches training using back propagation is repeated until a loss function converges to a first threshold value or until the count of training times is more than or equal to a second threshold value (i.e., when a preset stop condition is not met).)
LIU is not relied upon for teaching, but LI teaches: obtaining a codeword from a memory to create a first weight matrix of a neural network model; (LI [0036] teaches: "Aspects of the technology may include having all or some portions of a row vector of a matrix approximated by one or more codewords." LI [0037] teaches: "The first codeword W1 212, the second codeword W2 216, and the third codeword W3 220 may be stored in memory 230. Thus, the location of portions of the matrix 200 that may be approximated by one of the first codeword W1 212 (e.g., sub-vector 210, 218, and 224), the second codeword W2 216 (e.g., sub-vector 214), and the third codeword W3 220 (e.g., subvectors 222, 226, and 228) may reference the location in the memory 230 where the appropriate sub-vector is stored." LI [0006] teaches: "Using the codebook to index to the appropriate codeword (i.e., obtaining a codeword from memory) corresponding to the groups of sub-vectors, a small-footprint DNN matrix can be formed (i.e., to create a first weight matrix of a neural network model).")
determining, for a layer of the multi-layer NN model based on the codeword, that a weight matrix of the neural network model is the first weight matrix, and training the first weight matrix by using training data; (LI [0055] teaches: "The method begins with initialize codewords operation 502.” LI [0006] teaches: "Using the codebook to index to the appropriate codeword corresponding to the groups of sub-vectors, a small-footprint DNN matrix can be formed." LI [0006] teaches: "In aspects of technology, after the codebook is obtained, the codewords can be fine-tuned using a variety of neural network training techniques." LI [0044-0045] teaches: "The weight matrices
A
I
and the bias vectors
b
I
of the DNN can be estimated by minimizing the following cross entropy based loss function (i.e., training the first weight matrix):
L
X
t
r
=
-
∑
t
=
1
T
l
o
g
p
s
t
x
t
)
Where
X
t
r
=
x
1
,
:
∷
,
X
T
is a set of training feature vectors (i.e., by using training data) […].”)
Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of LIU and LI before them, to include LI's DNN matrix formation and training in LIU's neural network processing method. One would have been motivated to make such a combination in order to reduce the storage space required to store the DNN (LI [0027]).
Regarding Claim 2:
LIU in view of LI teaches the elements of claim 1 as outlined above. LIU further teaches:
The method according to claim 1, wherein when the first weight matrix is an initial weight matrix, the method further comprises: (LIU [Fig. 1B - leftmost matrix] teaches the initial weight matrix.)
dividing the initial weight matrix, to determine a codeword corresponding to the initial weight matrix. (LIU [0182] and [Fig. 1B] teaches grouping weights by clustering, dividing the grouped weights into same type groups to determine the codebook by using central weights.)
Regarding Claim 8:
LIU in view of LI teaches the elements of claim 1 as outlined above. LIU further teaches:
wherein the method further comprises: obtaining an index, wherein the index is a correspondence between the codeword and the weight matrix of the neural network model. (LIU [Fig. 1B] teaches determining a weight dictionary, where each value in the weight dictionary corresponds to a weight index and central weight (i.e., codeword) in the weight codebook.)
Regarding Claim 10:
The claim recites similar limitations as corresponding claim 1 and is rejected for similar reasons as claim 1 using similar teachings and rationale. Additionally, LIU teaches:
A neural network model training apparatus, wherein the device comprises a memory and a processor; the memory is configured to store instructions; and the processor is configured to execute the instructions in the memory, to perform the neural network model training method, wherein the method comprises: (LIU [0066] teaches: “An embodiment of the application, the computation unit is a neural network processor.” LIU [0172] teaches: “Another aspect of the application provides another electronic device, which may include a processor, a memory, a communication interface, and one or more programs, where the one or more programs are stored in the memory and are configured to be performed by the processor, and the programs include instructions configured for part or all of the operations described in the abovementioned image compression method.”)
a first obtaining unit, a first training unit, an updating unit, a storage unit, a second training unit, and a stop unit (LIU [0013] teaches a processor configured to perform the operations stored in a memory.)
Regarding Claim 11:
LIU in view of LI teaches the elements of claim 10 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 2 and is rejected for similar reasons as claim 2 using similar teachings and rationale. Furthermore, LIU teaches:
a division unit (LIU [0013] teaches a processor configured to perform the operations stored in a memory.)
Regarding Claim 17:
LIU in view of LI teaches the elements of claim 10 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 8 and is rejected for similar reasons as claim 8 using similar teachings and rationale. Furthermore, LIU teaches:
a second obtaining unit (LIU [0013] teaches a processor configured to perform the operations stored in a memory.)
Regarding Claim 19:
The claim recites similar limitations as corresponding claims 1 and 10 and is rejected for similar reasons as claims 1 and 10 using similar teachings and rationale. Additionally, LIU teaches:
A neural network model training device, wherein the device comprises a memory and a processor; the memory is configured to store instructions; and the processor is configured to execute the instructions in the memory, to perform the neural network model training method (LIU [0013] teaches a memory configured to store an operation instruction and processor configured to perform the operation stored in the memory.)
Claims 3 and 12 are rejected under 35 U.S.C. 103 as being unpatentable over LIU in view of LI as applied respectively above to claims 1 and 10, and further in view of CHENG (US20180247180A1), hereafter CHENG.
Regarding Claim 3:
LIU in view of LI teaches the elements of claim 2 as outlined above. LIU also teaches:
dividing the initial weight matrix into k submatrices […], wherein k is a positive integer greater than 1; (LIU [0005] teaches grouping and dividing the weight matrix (i.e., initial weight matrix) into m types, m being a positive integer (i.e., k is a positive integer). These divided weights form submatrices of the same type. LIU [0182] teaches dividing the weights into four types (i.e., greater than 1).
determining the n codewords as codewords corresponding to the initial weight matrix. ( LIU [Fig. 1B] teaches the weight codebook with central weights (i.e., codeword) that correspond to a weight matrix (i.e., initial weight matrix).)
However, LIU in view of LI is not relied upon for teaching, but CHENG teaches: dividing […] into k submatrices of a same dimension […] (CHENG [0019] and [0023] teaches uniformly dividing a matrix C into M groups, each group comprising of sub-vectors (i.e., into k submatrices), where M is a positive integer. This division process can be applied to LIU's leftmost matrix in Fig. 1B.)
LIU in view of CHENG teaches: performing clustering processing on the k submatrices of a same dimension, to obtain n codewords corresponding to the k submatrices of a same dimension, wherein n is a positive integer greater than 0, and n ≤ k; (LIU [0182] teaches performing clustering on the weight matrix to determine the weight codebook. The codeword can be understood as the central weights -1.3, -0.13, 0.23, and 1.50 (i.e., n codewords […] where in is a positive integer greater than 0, and n ≤ k) as shown in Fig. 1B. The combination of LIU and CHENG teaches the k submatrices being of a same dimension, as taught in CHENG [0019] and [0023].)
Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of LIU, LI, and CHENG before them, to include CHENG's uniform division of a matrix into groups in LIU and LI's neural network processing method. One would have been motivated to make such a combination in order to reduce the operational overhead and improve operation speed (CHENG [0050]).
Regarding Claim 12:
LIU in view of LI teaches the elements of claim 11 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 3 and is rejected for similar reasons as claim 3 using similar teachings and rationale. Furthermore, LIU teaches:
a first division subunit, a clustering subunit, and a first determining subunit (LIU [0013] teaches a processor configured to perform the operations stored in a memory.)
Claims 4 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over LIU in view of LI and CHENG as applied respectively above to claims 3 and 12, and further in view of CHOI (US 20180107926 A1), hereafter CHOI.
Regarding Claim 4:
LIU in view of LI and CHENG teaches the elements of claim 3 as outlined above. However, LIU in view of LI and CHENG is not relied upon for teaching, but CHOI teaches:
reducing each of the k submatrices of a same dimension into a one-dimensional vector, to obtain k one-dimensional vectors; (CHOI [0187-0189] and Equation (14) teaches vector quantization for constructing each n-dimensional vector:
v
i
=
W
n
i
-
1
+
1
W
n
i
-
1
+
2
…
W
n
i
T
Paragraph [0157] of the present application provides structure examples of the one-dimensional vectors as
[
a
1
,
a
2
,
a
3
,
a
4
,
a
5
,
a
6
]
. This shows a similar vector structure as CHOI's n-dimensional vectors. Therefore, under BRI, a one-dimensional vector can be understood as the n-dimensional vector constructed in CHOI. Furthermore, CHOI’s vector construction can be applied to LIU/CHENG's each group of sub-vectors (i.e., k-submatrices) in order to construct (i.e., obtain) a n-dimensional vectors.)
grouping the k one-dimensional vectors into n vector groups, wherein each vector group comprises at least one one-dimensional vector; (CHOI [0199] teaches grouping n-dimensional vectors into clusters.)
performing average calculation on element values at corresponding locations in all one-dimensional vectors that belong to an ith vector group in the k one-dimensional vectors, to obtain a codeword corresponding to all the one-dimensional vectors in the ith vector group, wherein i is an integer that ranges from 1 to n. (CHOI [0144] teaches calculating the mean, or the calculated Hessian-weighted mean, (i.e., performing average calculation) for member parameters (i.e., on element values at corresponding locations) of each cluster (i.e., ith vector group). One of ordinary skill in the art could apply this performing average calculation to LIU/CHENG's grouped subvectors.)
Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of LIU, LI, CHENG, and CHOI before them, to include CHOI's n-dimensional vector construction and mean calculation method in LIU, LI, and CHENG's neural network processing method. One would have been motivated to make such a combination in order to reduce memory requirements and minimize quantization loss (CHOI [0097]).
Regarding Claim 13:
LIU in view of LI and CHENG teaches the elements of claim 12 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 4 and is rejected for similar reasons as claim 4 using similar teachings and rationale. Furthermore, LIU teaches:
a dimension reduction subunit, a second grouping subunit, and a calculation subunit (LIU [0013] teaches a processor configured to perform the operations stored in a memory.)
Claims 5 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over LIU in view of LI as applied respectively above to claims 1 and 10, and further in view of KOTLER (US 20210342675 A1), hereafter KOTLER.
Regarding Claim 5:
LIU in view of LI teaches the elements of claim 1 as outlined above. LIU further teaches:
[…] weight matrix of the neural network model in the memory when the preset stop condition is not met […] (LIU [0174] teaches reading the weight data (i.e., weight matrix) from on-chip cache (i.e., memory) to improve the performance of the neural network (i.e., neural network model) training, such as repeating back propagation until a loss function converges to a first threshold value or until the count of training times is more than or equal to a second threshold value (i.e., when a preset stop condition is not met).)
However, LIU in view of LI is not relied upon for teaching, but KOTLER teaches: releasing the weight matrix of the neural network model in the memory when the preset condition is not met. (KOTLER [0049] teaches releasing the machine learning network (MLN) (i.e., neural network model) weights from memory (i.e., releasing the weight matrix of the neural network model in memory) once they are no longer needed. For example, if a computation in a certain time period does not depend on the weights, these weights can be released from memory.)
Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of LIU, LI, and KOTLER before them, to include KOTLER's releasing of weights in LIU and LI's neural network processing method. One would have been motivated to make such a combination in order to free up space for other computations and reduce the overall memory usage (KOTLER [0045] and [0048]).
Regarding Claim 14:
LIU in view of LI teaches the elements of claim 10 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 5 and is rejected for similar reasons as claim 5 using similar teachings and rationale. Furthermore, LIU teaches:
a releasing unit (LIU [0013] teaches a processor configured to perform the operations stored in a memory.)
Claims 6-7, 9, 15-16, and 18 are rejected under 35 U.S.C. 103 as being unpatentable over LIU in view of LI as applied respectively above to claims 1 and 10, and further in view of CHOI (US 20180107926 A1), hereafter CHOI.
Regarding Claim 6:
LIU in view of LI teaches the elements of claim 1 as outlined above. LIU further teaches:
determine a weight gradient […] when the preset stop condition is not met; (LIU [0611] teaches computing weight gradients for reverse training (updating) the weights, and training using back propagation is repeated until a loss function converges to a first threshold value or until the count of training times is more than or equal to a second threshold value (i.e., when a preset stop condition is not met), and updating the target model (i.e., weights of the model) according to the loss function. The target model is the compression neural network model that has been compressed using the weight codebook method, and has an initial weight matrix (i.e., first weight matrix) prior to training.)
However, LIU in view of LI is not relied upon for teaching, but CHOI teaches: determining a weight gradient of the first weight matrix of the neural network model […] (CHOI [0210] teaches equation (16) that calculates (i.e., determining) the average gradient of the network loss function with respect to network parameters (i.e., weights). Specifically, the gradient term
∑
w
l
∈
C
i
,
j
∂
L
(
w
)
∂
w
l
w
=
w
-
t
-
1
is shown in equation (16)(c):
g
i
,
j
t
-
1
=
1
C
i
,
j
∑
w
l
∈
C
i
,
j
∂
L
(
w
)
∂
w
l
w
=
w
-
t
-
1
This calculation can be applied to LIU's initial weight matrix (i.e., first weight matrix).
LIU in view of CHOI also teaches: determining a codeword gradient based on the first weight gradient, and determining the updated codeword based on the codeword gradient. (CHOI [0210] teaches equation (16)(c) that determines
g
i
,
j
. Under BRI, a codeword gradient can be understood as
g
i
,
j
, which is determined based on the gradient of the initial weight matrix (i.e., LIU’s initial weight matrix). CHOI [0209] teaches determining an updated codeword by using equation (16)(b), where
c
i
,
j
t
represents the shared quantized vector (i.e., codeword) for cluster i.)
Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of LIU, LI, and CHOI before them, to include CHOI's weight gradient determination in LIU and LI's neural network processing method. One would have been motivated to make such a combination in order to reduce memory requirements and minimize quantization loss (CHOI [0097]).
Regarding Claim 7:
LIU in view of LI and CHOI teaches the elements of claim 6 as outlined above. CHOI also further teaches:
performing weighted summation on weight gradients that are in the weight gradient and that are of submatrices corresponding to index numbers that belong to a jth codeword, to obtain a codeword gradient corresponding to the jth codeword, wherein j is an integer that ranges from 1 to n; (see CHOI [0210] equation (16)(c). Additionally, CHOI [0208] teaches j being
1
≤
j
≤
n
.)
optimizing the codeword gradient corresponding to the jth codeword, to obtain an update amount of the jth codeword; (see CHOI [0209] equation (16)(b))
updating the jth codeword by using the update amount of the jth codeword, to obtain an updated jth codeword. (see CHOI [0209] equation (16)(b))
Regarding Claim 9:
LIU in view of LI teaches the elements of claim 1 as outlined above. However, LIU in view of LI is not relied upon for teaching, but LIU in view of LI and CHOI teaches:
wherein the preset stop condition comprises one or more of the following conditions: (LIU [0611] teaches a preset stop condition)
an output value of a loss function used by the neural network model reaches a preset threshold, wherein the loss function is used to measure the difference between the result output by the neural network model for the training data and the result label value corresponding to the training data. (CHOI [0065] teaches adjusting parameters until the objective function of the stochastic gradient descent stops decreasing (i.e., an output value of a loss function used by the neural network model reaches a preset threshold). CHOI [0004] describes measuring the error of the objective function with the goal of the machine learning system (i.e., network model) correctly labeling the categories in the image training data.)
Accordingly, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of LIU, LI, and CHOI before them, to include CHOI's adjusting parameters until the objective function of the stochastic gradient descent stops decreasing in LIU and LI's neural network processing method. One would have been motivated to make such a combination in order to reduce memory requirements and minimize quantization loss (CHOI [0097]).
Regarding Claim 15:
LIU in view of LI teaches the elements of claim 10 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 6 and is rejected for similar reasons as claim 6 using similar teachings and rationale. Furthermore, LIU teaches:
a second determining subunit and a third determining subunit (LIU [0013] teaches a processor configured to perform the operations stored in a memory.)
Regarding Claim 16:
LIU in view of LI and CHOI teaches the elements of claim 15 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 7 and is rejected for similar reasons as claim 7 using similar teachings and rationale. Furthermore, LIU teaches:
a first obtaining subunit¸ a second obtaining subunit, and a third obtaining subunit (LIU [0013] teaches a processor configured to perform the operations stored in a memory.)
Regarding Claim 18:
LIU in view of LI teaches the elements of claim 10 as outlined above. Additionally, the claim recites similar limitations as corresponding claim 9 and is rejected for similar reasons as claim 9 using similar teachings and rationale.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).D!!!!!!)
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Alvaro S Laham Bauzo whose telephone number is (571)272-5650. The examiner can normally be reached Mon-Fri 7:30 AM - 11:00 AM | 1:00 PM - 5:30 PM ET.
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, Usmaan Saeed can be reached on (571) 272-4046. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/A.S.L./Examiner, Art Unit 2146
/USMAAN SAEED/Supervisory Patent Examiner, Art Unit 2146
Prosecution Insights