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 .
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 1-18 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Claim 1 recites a method, one of the four statutory categories of patentable subject matter. However, the claim recites the steps of calculating spectral information associated with a target graph (a mental or mathematical process, dependent upon the particular information calculated); and generating an embedding representation of the target graph based on information on nodes constituting the target graph and the spectral information (a mental or mathematical process). Thus the claim recites the abstract idea of generating an embedding representation of a graph based on spectral information of the graph.
The claim does not include any additional elements which could integrate the abstract idea into a practical application, because the only additional element recited is at least one processing device upon which to perform the method’s steps. Performance of an abstract idea on generic computer components cannot integrate an abstract idea into a practical application (see MPEP 2106.05(f)(2)). Therefore, the claim is directed to the abstract idea of generating an embedding representation of a graph based on spectral information of the graph.
Further, implementation on generic computer components cannot represent significantly more than the abstract idea itself (MPEP 2106.05(f)(2)), so the claim is subject matter ineligible.
Claim 2-7, dependent upon Claim 1, each only recite additional steps of the abstract idea, each of which is a mental or mathematical process (Claim 2: derive an ego graph, calculate spectral information (mental); Claim 3: calculate eigenvalue and eigenvector as the spectral information (mathematical); Claim 4: calculate angle information as the spectral information (mental); Claim 5: derive an ego graph, derive a sub graph, calculate spectral information (mental); Claim 6: perform label updating according to WL algorithm (mathematical); generate the embedding (mental); Claim 7: determining by comparing representations (mental)), but no new additional elements, thus no additional elements which could integrate the abstract idea into a practical application nor provide significantly more than the abstract idea itself.
Claims 8 and 16, dependent upon Claim 1, each only recite additional steps of the abstract idea (Claim 8: aggregating information; generate the embedding representation are both mental processes; Claim 16: generating an intermediate embedding representation; aggregating information) which are specifically recited as being performed by a graph neural network. The performance of abstract idea steps by using a computer or other machinery (i.e. a neural network) as a tool cannot integrate the abstract idea into a practical application nor provide significantly more than the abstract idea itself (MPEP 2106.05(f)(2)).
Similarly, Claim 9, dependent upon Claim 8, only recites additional steps of the abstract idea (aggregating information which is in the form of a set of a plurality of elements; transforming information to a form of vector or matrix, and aggregating information are all mental processes) which are specifically recited as being performed by a neural network. The performance of abstract idea steps by using a computer or other machinery (i.e. a neural network) as a tool cannot integrate the abstract idea into a practical application nor provide significantly more than the abstract idea itself (MPEP 2106.05(f)(2)).
Claims 10-14 recite only additional mental process steps of the abstract idea (Claim 10: changing a value, aggregating information; Claim 11: changing a value to an irrational number; Claim 12: predicting a label, updating the value; Claims 13-14: performing multiplication, addition, concatenation), but no new additional elements, thus no additional elements which could integrate the abstract idea into a practical application nor provide significantly more than the abstract idea itself.
Claim 15, dependent upon Claim 10, merely reiterates that Claim 10 recites to use “a computer or other machinery as a tool” to perform the mental process step of Claim 10 (i.e. the GNN is configured to perform the process that it performs), which by MPEP 2106.05(f)(2) cannot integrate the abstract idea into a practical application nor provide significantly more than the abstract idea itself.
Claims 17 and 18 recite, respectively, one or more processors and memory; or a non-transitory computer readable medium, with which to perform the abstract idea steps of Claim 1. As performance of abstract idea steps by using a generic computer components cannot integrate the abstract idea into a practical application nor provide significantly more than the abstract idea itself (MPEP 2106.05(f)(2)), Claims 17 and 18 are rejected for reasons set forth in the rejection of Claim 1.
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.
Claims 1, 2, 5, and 8-18 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Zhu et al., “Transfer Learning of Graph Neural Networks with Ego-graph Information Maximization,” (as provided by the applicant in the Information Disclosure Statement dated 7/2/2024).
Regarding Claim 1, Zhu teaches a method for processing a graph (Zhu, title, “Transfer Learning on Graph Neural Networks) performed by at least one computing device (Zhu, pg. 1, Footnote 1, “Code and processed data are available” denotes that the method is performed on a computer), the method comprising: calculating spectral information associated with a target graph (Zhu, pg. 2, Fig. 1 & pg. 8, 3rd paragraph, node features are spectral embeddings, “spectral and other pre-computed node embeddings are also applicable” to use as the input node features); generating an embedding representation of the target graph based on information on nodes constituting the target graph and the spectral information (Zhu, pg. 4, Fig. 2, and following text, “We aim to train a GNN encoder
Ψ
“ which acts on graph
g
i
and node features/spectral information
x
i
to generate embedding representation of the target graph
z
i
).
Regarding Claim 2, Zhu teaches the method of Claim 1 (and thus the rejection of Claim 1 is incorporated). Zhu further teaches deriving an ego-graph for at least some of the nodes in the target graph (Zhu, pg. 4, Fig. 2 illustrates the sub-graph ego-graph); and calculating the spectral information of the ego-graph (Zhu, pg. 4, Fig. 2, and following text, “
(
g
i
,
x
i
)
” where the
x
i
is the node features/spectral embedding/spectral information of the ego-graph).
Regarding Claim 5, Zhu teaches the method of Claim 2 (and thus the rejection of Claim 2 is incorporated). Zhu further teaches deriving an ego-graph for at least some of the nodes from the target graph (Zhu, pg. 4, Fig. 2); deriving a sub-graph from which an ego node is removed from the ego-graph; and calculating spectral information of the derived sub-graph (Zhu, pg. 5, 2nd paragraph, “while
Φ
outputs representation of the neighbor nodes” i.e. of the subgraph without the center node, also see pg. 15, expression in the middle of the page for the operation of the graph neural network to compute
z
i
operates on the sub-graph of
N
(
x
i
) i.e. the neighbors, which do not include the center node).
Regarding Claim 8, Zhu teaches the method of Claim 1 (and thus the rejection of Claim 1 is incorporated). Zhu further teaches aggregating the information of the nodes and the spectral information; and generating the embedding representation by inputting the aggregated information into a graph neural network (GNN) (Zhu, pg. 4, Fig. 2, and following text, “We aim to train a GNN encoder
Ψ
” which has as input aggregated
(
g
i
,
x
i
)
to get “node embedding
z
i
=
Ψ
(
g
i
,
x
i
)
”).
Regarding Claim 9, Zhu teaches the method of Claim 8 (and thus the rejection of Claim 8 is incorporated). Zhu further teaches wherein the spectral information is information in a form of a multi-set comprising a plurality of spectral elements (Zhu, pg. 4, Fig. 2, and following text,
g
i
,
x
i
is a multi-set)
g
i
,
x
i
; and the aggregating of the information of the nodes and the spectral information comprises: transforming of the spectral information through a neural network module that transforms the data in a form of a multi-set into data in a form of a vector or matrix (Zhu, pg. 15, the expression in the middle of the page for the operation of the graph neural network to compute
z
i
transforms the node features from the multi-set into vectors).
Regarding Claim 10, Zhu teaches the method of Claim 8 (and thus the rejection of Claim 8 is incorporated). Zhu further teaches changing a value of any one of the information of the nodes and the spectral information by reflecting a specific value to any one of the information of the nodes and the spectral information; and aggregating any changed information and other information (Zhu, pg. 15, the expression in the middle of the page for the operation of the graph neural network to compute
z
i
reflects the information of the node spectral features into the node embeddings).
Regarding Claim 11, Zhu teaches the method of Claim 10 (and thus the rejection of Claim 10 is incorporated). Zhu further teaches wherein the specific value is an irrational number (Zhu, pg. 15, the expression for
Z
(
l
)
which is what is computed during the learning/reflection includes a square root of a matrix, which for general graphs will be irrational).
Regarding Claim 12, Zhu teaches the method of Claim 10 (and thus the rejection of Claim 10 is incorporated). Zhu further teaches wherein the specific value is a value based on a learnable parameter (the encoding neural network is trained and this includes learnable parameters that determine the encoding
Ψ
) and the method further comprises: predicting a label for a predetermined task based on the generated embedding representation; and updating the value of the learnable parameter based on a difference between the predicted label and the correct label (Zhu, pg. 4, Fig. 2, where the task is to “compute the probability of an edge e belongs to the given ego-graph” and the encoder is trained based upon the discriminator’s prediction, see loss function of Eq. (1)).
Regarding Claim 13, Zhu teaches the method of Claim 10 (and thus the rejection of Claim 10 is incorporated). Zhu further teaches wherein the reflecting of the specific value is performed based on a multiplication operation and the aggregating of any changed information and other information is performed based on an addition operation (Zhu, pg. 15, the expression in the middle of the page for the operation of the graph neural network to compute
z
i
to perform the reflecting includes both addition and multiplication operations).
Regarding Claim 14, Zhu teaches the method of Claim 8 (and thus the rejection of Claim 8 is incorporated). Zhu further teaches wherein the aggregating is performed based on a concatenation operation (Zhu, pg. 15, 2nd paragraph, “in the i-th layer [we perform a] concatenation operation”).
Regarding Claim 15, Zhu teaches the method of Claim 8 (and thus the rejection of Claim 8 is incorporated). Zhu further teaches wherein the GNN is a neural network configured to generate an embedding representation of a graph by aggregating information of neighboring nodes (Zhu, pg. 15, the expression in the middle of the page for the operation of the graph neural network to compute
z
i
aggregates by summation over a set of neighboring nodes).
Regarding Claim 16, Zhu teaches the method of Claim 8 (and thus the rejection of Claim 8 is incorporated). Zhu further teaches generating an intermediate representation of the target graph by inputting the information of the nodes into a graph neural network (Zhu, pg. 15, the expression in the middle of the page for the operation of the graph neural network to compute
z
i
computes intermediate representations
z
i
(
k
)
by inputting information
x
i
); and generating the embedding representation by aggregating the intermediate embedding representation and the spectral information (Zhu, pg. 15, the expression in the middle of the page for the operation of the graph neural network aggregates
z
j
(
k
-
1
)
and the node features).
Claim 17 recites a system … comprising: one or more processors; and memory to perform precisely the method of Claim 1. As Zhu performs their method on a computer (Zhu, pg. 1, Footnote 1, “Code and processed data are available”), in which a processor and memory are inherent, Claim 17 is rejected for reasons set forth in the rejection of Claim 1. Similarly, Claim 18 recites a non-transitory computer readable storage medium to perform the method of Claim 1, inherent in Zhu’s computer implementation, and is also rejected for reasons set forth in the rejection of Claim 1.
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.
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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Zhu, in view of Feng et al., “Representation Learning for Scale-Free Networks.”
Regarding Claim 3, Zhu teaches the method of Claim 2 (and thus the rejection of Claim 2 is incorporated). Zhu teaches that the node features, identified as the claimed spectral information, can be spectral embeddings, but does not explicitly teach that these spectral embeddings be at least one of an eigenvalue and an eigenvector of the ego graph. However, Feng teaches that spectral embeddings can be determined by calculating eigenvectors (Feng, pg. 4, 2nd column, “Degree Penalty based Spectral Embeddings” are calculated using “an eigenvector of L” with Eqs. (11-13).) It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to calculate the input node features/spectral embeddings of Zhu using the DP-Spectral method of Feng. The motivation to do so is Feng’s embeddings “outperform state-of-the-art embedding models in various network mining tasks” (Feng, Abstract), i.e. are useful node features.
Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over Zhu, in view of Hsu et al., “Efficient and interpretable graph network representation for angle-dependent properties applied to optical spectroscopy.”
Regarding Claim 4, Zhu teaches the method of Claim 2 (and thus the rejection of Claim 2 is incorporated). Zhu teaches that the node features, identified as the claimed spectral information, can be spectral embeddings, but does not explicitly teach that these spectral embeddings be angle information of the ego graph. However, Hsu, also in the analogous art of graph representations, teaches that node features can include angle information (Hsu, pg. 7, 1st column, “ALIGNN-d representation … The nodes and edges in L(G) and L’(G), on the other hand, represent bonds and angles”). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to additionally include, in the node features of Zhu, features like the angle information of Hsu. The motivation to do so is to use the method of Zhu to obtain representations of graphs in the spectroscopy area, as does Hsu, where “This simple extension leads to a memory-efficient graph representation that captures the complete geometry of atomic structures” (Hsu, Abstract) – that is, angle information is important information to capture when applying graph representation methods, such as those to Zhu, to application areas such as chemistry and spectroscopy.
Claims 6 and 7 are rejected under 35 U.S.C. 103 as being unpatentable over Zhu, in view of Verma et al., “Learning Universal Graph Neural Network Embeddings with Aid of Transfer Learning,” (as provided by the applicant in the Information Disclosure Statement of 7/2/2024).
Regarding Claim 6, Zhu teaches the method of Claim 1 (and thus the rejection of Claim 1 is incorporated). Zhu further teaches performing repetitive node label updating … based on the information on the nodes and the spectral information; and generating the embedding representation based on a final label of the nodes obtained through the node label updating (Zhu, pg. 15, the expression in the middle of the page for the operation of the graph neural network to compute
z
i
includes node label updating over k iterations based at the label at the previous iteration). Zhu does not teach using a Weisfeiler-Lehmann algorithm. However, Verma, in the analogous art of graph representation transfer learning, teaches this limitation (Verma, pg. 4, Fig. 2, “WL Kernel” in the task decoder, analogous to the discriminator of Zhu). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use a WL algorithm, such as the one of Verma, in the discriminator of Zhu, in order to learn the encoder of Zhu. The motivation to do so is to apply the transfer learning of Zhu to the graph isomorphism multi-task problems of Verma.
Regarding Claim 7, the Zhu/Verma combination of Claim 6 teaches the method of Claim 6 (and thus the rejection of Claim 6 is incorporated). Zhu further teaches wherein the target graph comprises a first graph and a second graph, and the method further comprises: determining whether the first graph and the second graph are isomorphic by comparing an embedding representation of the first graph and an embedding representation of the second graph (Zhu, pg. 4, Fig. 2 where the “Discriminator” compares the two graphs to see whether the predictions of the edge existences match, i.e. whether the graphs are isomorphic, also see pg. 2, 2nd column, “structural-equivalent role identification” & pg. 8, Table 1, “identifying structural equivalent nodes”).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure: Epasto, US PG Pub 2020/0035002, also teaches a different method of removing the center node in an ego-graph for graph representation learning (Epasto, [0046], “in some embodiments, the ego-net of u may not include the node u itself’).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BRIAN M SMITH whose telephone number is (469)295-9104. The examiner can normally be reached Monday - Friday, 8:00am - 4pm Pacific.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Kakali Chaki can be reached at (571) 272-3719. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of 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.
/BRIAN M SMITH/Primary Examiner, Art Unit 2122
Prosecution Insights