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 .
Response to Amendment
Applicants' arguments and remarks filed on 07/23/2026 have been fully considered.
Applicants' amendments overcome objections to the specification.
Applicants' amendments overcome objections to the claims.
Applicants' amendments overcome the previous 35 U.S.C. 112(b) rejection.
Claims 1-5 were previously canceled through Preliminary Amendment filed 07/26/2024.
Claims 6-8 are currently amended.
Claims 6-9 are pending.
Response to Arguments
Applicant's arguments filed 07/23/2026 have been fully considered but they are not persuasive. To the extent the arguments are directed to limitations newly added by amendment, they have been addressed in the updated grounds of rejection below.
Applicant argues that the combination of Li et al. (US 2022/0163649 A1), Chen et al. (US 2023/0046274 A1), and Garimella et al. (US 2022/0274625 A1) does not disclose or suggest “searching … for anomalies and patterns in features of the graph structure, wherein the anomalies searched for include associated attributes of a first object of the plurality of objects compared with other objects of the plurality of objects that surround the first object” (Remarks, p. 6). The examiner respectfully disagrees. Li et al. (‘649) searches for exactly this type of anomaly. Li et al. (‘649) compares the range attribute of object Ox against the ranges of the surrounding objects O1 and O2 (“comparing, at the hardware processor 112, whether O1O2=2R03′-R01-R02”; [0094]). Li et al. (‘649) then compares the radar cross-section and lifetime attributes of one dynamic object against the other dynamic object in the same group of objects (“finding the dynamic object with less RCS or lifetime among the two dynamic objects and increasing the corresponding ghost probability”; [0095]). All of these objects are located in the surrounding area of the vehicle ([0089]). Garimella et al. (‘625) in turn teaches that the graph structure represents each object’s attributes as compared with every other connected object, because each edge “stores one or more relative offset values representing the differences in attribute values between the pair of nodes connected by the edge” ([0022]), and the graph “may be fully connected, so that a distinct edge is associated with each different pair of nodes” ([0021]). In the combination, Li et al. (‘649)’s attribute comparisons are performed on these node and edge features. See the rejection of claim 6 below.
Applicant argues that Li et al. (‘649) identifies ghost objects based on measurements of a single sensor and does not disclose or suggest a graph structure (Remarks, p. 7). The examiner respectfully disagrees with the characterization of Li et al. (‘649), which receives “sensor data 124 from one or more sensors 122” ([0089]). In any event, the rejection does not rely on Li et al. (‘649) for the multi-sensor limitations or for the graph structure. Chen et al. (‘274) is relied upon for generating object hypotheses using several individual sensors and for the cross-sensor comparison, and Garimella et al. (‘625) is relied upon for the graph structure. In response to applicant’s arguments against the references individually, one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986).
Applicant argues that Chen et al. (‘274)’s machine-learning models do not search a graph structure for anomalies and patterns (Remarks, p. 7). Chen et al. (‘274) is not relied upon to teach searching a graph structure. The argument attacks Chen et al. (‘274) individually and is not persuasive for the reasons given in paragraph 11 above.
Applicant argues that Garimella et al. (‘625) generates a neural network with map and entity nodes and edges and uses it to predict the state of an entity, which does not involve searching a graph structure for anomalies and patterns (Remarks, p. 7). The rejection does not rely on Garimella et al. (‘625) to perform the anomaly and pattern search. Li et al. (‘649) teaches the search for anomalies and patterns and the ghost object classification. Garimella et al. (‘625) is relied upon for storing object attributes, inter-object relational information, and map data as the nodes and edges of a graph structure, the features of which are searched in the combined system. The rejection below has been clarified to make this attribution explicit.
Applicant further argues that Garimella et al. (‘625) does not disclose or suggest that its neural network has any use in connection with ghost object detection (Remarks, p. 7). A secondary reference need not address the same problem as the applicant or as the primary reference. “[A]ny need or problem known in the field of endeavor at the time of invention and addressed by the patent can provide a reason for combining the elements in the manner claimed.” KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 420, 82 USPQ2d 1385 (2007). Garimella et al. (‘625) is analogous art because it is from the same field of endeavor as the claimed invention, namely modeling a vehicle’s perceived environment from on-board sensor data and map data ([0013], [0045], [0051]). See In re Bigio, 381 F.3d 1320, 1325, 72 USPQ2d 1209 (Fed. Cir. 2004); MPEP § 2141.01(a). The examiner also notes that the remarks state that the person of ordinary skill “would have found it obvious” to use Garimella’s neural network to identify anomalies used for ghost classification (Remarks, p. 7). The examiner treats this statement as a typographical error and does not rely on it.
Each reason to combine relied upon below is drawn from the references’ own stated problems and advantages. Li et al. (‘649) identifies a need for a general ghost detection solution suited to the larger number of targets in automobile radar tracking ([0005-0006]). Chen et al. (‘274) identifies fast and reliable verification of radar detections ([0019]). Garimella et al. (‘625) identifies the storage of relative attribute differences between objects and single-pass processing of the environment as a whole ([0011], [0022], [0029]).
Applicant’s remaining arguments regarding claims 7-9 rely on their dependency from claim 6 (Remarks, p. 7). Because claim 6 remains rejected, claims 7-9 remain rejected for the reasons set forth below.
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.
Claims 6-9 are rejected under 35 U.S.C. 103 as being unpatentable over Li et al. (US 2022/0163649 A1) in view of Chen et al. (US 2023/0046274 A1) and further in view of Garimella et al. (US 2022/0274625 A1).
Regarding Claim 6, Li et al. (‘649) in view of Chen et al. (‘274) and Garimella et al. (‘625) teaches:
Li et al. (‘649) teaches: a method for detecting ghost objects in sensor measurements of an environment of a vehicle, the method comprising: Li et al. (‘649) is directed to detecting ghost objects arising from multipath reflections in sensor data of a vehicle’s surroundings ([0003]: “There may be strong reflective surfaces and targets in the vicinity of the vehicle resulting in the detection of ghost objects”). Li et al. (‘649) receives ([0089]: “sensor data 124 from one or more sensors 122 in communication with the hardware processor 112 and positioned such that the surrounding area 10 is within a field of view of the one or more sensors 122”).
Li et al. (‘649) does not explicitly teach obtaining map context information from a digital road map, but Chen et al. (‘274) teaches obtaining map context information from a digital road map: Chen et al. (‘274) teaches that the autonomous vehicle determines its location with respect to ([0026]: “fixed objects of the driving environment 101 (e.g. roadways, lane boundaries, intersections, sidewalks, crosswalks, road signs, curbs, surrounding buildings, etc.) whose locations can be provided by map information 135”).
Li et al. (‘649) teaches: generating … object hypotheses of a plurality of objects having associated attributes in the environment of the vehicle: Li et al. (‘649) teaches ([0089]: “detecting, at the hardware processor 112, one or more objects 102, 102a-n, 104, 104a-n from the sensor data 124”). Each detected object is an object hypothesis, since Li et al. (‘649) subsequently determines whether it is a real object or a ghost object. The detected objects have associated attributes, including azimuth angle, range (R01, R02, R03’), RCS, and lifetime ([0090-0091], [0095]).
Li et al. (‘649) does not explicitly teach that the object hypotheses are generated using several individual sensor measurements by several individual sensors of the vehicle, but Chen et al. (‘274) teaches this limitation: Chen et al. (‘274) teaches a sensing system that ([0021]: “can include a radar 114 (or multiple radars 114)” and “can include a lidar 112”), and that ([0023]: “can further include one or more cameras 118 to capture images of the driving environment 101”). Chen et al. (‘274)’s perception system ([0024]: “can be configured to detect and track objects in the driving environment 101 and to recognize the detected objects” and “can use radar data in combination with the data captured by the camera(s) 118”).
It would have been obvious to a POSITA before the effective filing date of the claimed invention to incorporate the map information and multi-sensor object detection of Chen et al. (‘274) into the ghost object detection method of Li et al. (‘649). The motivation is found in the references themselves. Li et al. (‘649) expressly recognizes that prior ghost detection approaches are ([0005]: “application-specified and lacking a general solution” and that ([0005]: “there are chances to identify real object as ghost objects since there are more targets in automobile radar tracking”). Li et al. (‘649) further acknowledges that ([0004]: “road objects and ghost area are used to assist in identifying ghost objects”). Chen et al. (‘274) teaches that its system is capable of ([0019]: “combining radar data with camera data for fast and reliable verification of radar detections”) and uses map information 135 to locate objects relative to fixed road features ([0026]). A POSITA would have been motivated by Chen et al. (‘274)’s stated advantage of reliable radar verification to address the misclassification risk that Li et al. (‘649) itself identifies. The reasonable expectation of success is grounded in Chen et al. (‘274)’s demonstrated implementation of map-informed, multi-sensor perception in an autonomous vehicle.
Li et al. (‘649) teaches: generating social context information about the plurality of objects of the object hypotheses in relation to one another, wherein the social context information is generated from relationships between the objects of the object hypotheses located in the vicinity of the vehicle: Li et al. (‘649) determines the relationships between the detected objects in the surrounding area of the vehicle, including their collinearity with the radar ([0090]: “verifying, at the hardware processor 112, that two objects and radar are colinear by checking azimuth angle”), the inter-object distances ([0091]: “determining, at the hardware processor 112, the distance from O1 to O2” and “the distance from O2 to Ox”), and the triangle relationship among the objects ([0085-0088]: “We can always form a triangle having two edges with the same length”).
Li et al. (‘649) and Chen et al. (‘274) do not explicitly teach storing the obtained map context information, the object hypotheses and associated attributes, and the social context information in a graph structure comprising nodes and edges, wherein relational information is depicted in the graph structure by the edges, but Garimella et al. (‘625) teaches this limitation: Garimella et al. (‘625) teaches that the GNN component stores ([0045]: “vectorized representations of map elements within the map(s) 224, and entities or other objects perceived by the perception component 222 … as nodes in a graph structure”). The nodes ([0022]: “may correspond to a native data network that stores the intrinsic attributes associated with the various objects in the environment”), and ([0021]: “edge data (or edge features) associated with the edge may include relative data between the pair of nodes”). Garimella et al. (‘625) receives map data including ([0013]: “topologies, intersections, streets, roads, terrain, and the environment in general”) and stores the map elements as nodes ([0014], [0045]). In the combined system, Chen et al. (‘274)’s map information 135 is stored as map element nodes, Li et al. (‘649)’s object hypotheses and their associated attributes are stored as entity nodes, and Li et al. (‘649)’s inter-object relationships are stored as edge features.
It would have been obvious to a POSITA before the effective filing date of the claimed invention to store the data of the combined system of Li et al. (‘649) and Chen et al. (‘274) in the graph structure of Garimella et al. (‘625). Garimella et al. (‘625) itself provides the motivation. Garimella et al. (‘625) teaches that its edges store ([0022]: “one or more relative offset values representing the differences in attribute values between the pair of nodes connected by the edge”), which are the same kind of relative quantities, such as relative distance and relative range, that Li et al. (‘649) computes as the basis of its ghost detection. Garimella et al. (‘625) further teaches that ([0029]: “an execution of a single inference process (or operation) on the GNN may output an inferred updated representation of the environment as a whole”), and that its techniques provide ([0011]: “technical advantages in reducing computing resources and improving efficiency”). These advantages address the concern Li et al. (‘649) itself raises regarding the larger number of targets in automobile radar tracking ([0005]). This motivation arises from the references’ own teachings and not from Applicant’s disclosure. The reasonable expectation of success is grounded in Garimella et al. (‘625)’s demonstrated implementation of a graph structure populated from multi-sensor perception and map data in an autonomous vehicle ([0045], [0051]), and in the fact that Li et al. (‘649)’s operations are arithmetic comparisons of quantities that such a graph structure stores.
Li et al. (‘649) teaches: searching … for anomalies and patterns … wherein the anomalies searched for include associated attributes of a first object of the plurality of objects compared with other objects of the plurality of objects that surround the first object: Li et al. (‘649) searches for patterns by testing whether the relationships among the detected objects satisfy the geometric conditions of known multipath reflection cases ([0093]: “comparing, at the hardware processor 112, whether O1O2=0.5O2O3′”). Li et al. (‘649) compares the range attribute of object Ox, the first object, with the ranges of the other objects O1 and O2 ([0094]: “comparing, at the hardware processor 112, whether O1O2=2R03′-R01-R02”). Li et al. (‘649) then searches for an anomalous attribute among the objects by comparing the radar cross-section and lifetime of one dynamic object with the other ([0095]: “finding the dynamic object with less RCS or lifetime among the two dynamic objects and increasing the corresponding ghost probability”). All of these objects are located in the surrounding area 10 of the vehicle ([0089]) and thus surround one another.
Li et al. (‘649) does not explicitly teach performing this search in features of the graph structure or based on the map context information, the social context information, and the object hypotheses and the associated attributes, but Garimella et al. (‘625) teaches this limitation: Garimella et al. (‘625) teaches that each object’s attributes are represented in the graph as differences from every connected object, because each edge ([0022]: “stores one or more relative offset values representing the differences in attribute values between the pair of nodes”) in a graph that ([0021]: “may be fully connected, so that a distinct edge is associated with each different pair of nodes”). Garimella et al. (‘625) further teaches that edges to nearby objects are emphasized, in that ([0023]: “the edge connecting the entity object to another object in the environment that is closer, dynamic, and/or in front of the entity, may be assigned a higher weight value”). Garimella et al. (‘625) also teaches edges between entity nodes and map element nodes that store ([0021]: “the relative distance, relative yaw, relative velocity, relative pose, relative size, relative acceleration, relative permissibility, and the like, between the first entity node and the second map element node”), such that the features of the graph structure include map context information. In the combined system, Li et al. (‘649)’s pattern and anomaly tests are performed on the node attributes (the object hypotheses and associated attributes), the entity-to-entity edge features (the social context information), and the entity-to-map-element edge features (the map context information) of the graph structure. The motivation and reasonable expectation of success are the same as those set forth above for incorporating Garimella et al. (‘625).
Li et al. (‘649) teaches: classifying one or more of the plurality of objects as ghost objects using detected anomalies and patterns in the graph structure: Li et al. (‘649) teaches ([0096]: “comparing, at the hardware processor 112, whether the ghost probability of the less confidence object is higher than a upper threshold, and marking it as ghost object”). In the combined system, the anomalies and patterns used for this classification are those detected in the features of the graph structure of Garimella et al. (‘625), as set forth above.
Li et al. (‘649) does not explicitly teach comparing a detection of ghost objects in further sensor measurements by at least one further sensor with at least one object hypothesis of at least one sensor measurement of the several individual sensors to determine whether the detection of ghost objects in the further sensor measurements is plausible or is implausible, but Chen et al. (‘274) teaches this limitation: In Chen et al. (‘274), radar 114 is the at least one further sensor, and its radar return points represent ([0025]: “a list of potential real or spurious object”). Camera 118 is one of the several individual sensors. Chen et al. (‘274) teaches that the object tracking module segments ([0036]: “a patch of the acquired image with the location that approximately corresponds to the tracked radar blob’s bounding box” and provides the radar data and the camera patch “to RVM 132 for validation”). RVM 132 ([0025]: “can include one or more trained models that are used to process some or all of the above data to classify the radar returns as genuine or spurious”). The outcome is a plausibility determination: if the returned probability is at or above a first threshold p1, the tracking module ([0037]: “can determine with high confidence that the blob being tracked is still associated with the real object,” and if it is below a second threshold p2, the module “can conclude that the blob being tracked is not associated with a real object”). Under the broadest reasonable interpretation, jointly evaluating co-located radar data and the camera-based object hypothesis to output a real-object likelihood constitutes comparing the radar detection with the camera-based object hypothesis. In the combined system, Li et al. (‘649)’s ghost object classification of the radar-measured objects is compared with the camera-based object hypothesis at the same location to determine whether the ghost classification is plausible or implausible. The motivation and reasonable expectation of success are the same as those set forth above for incorporating Chen et al. (‘274), namely fast and reliable verification of radar detections ([0019]).
It would have been obvious to a POSITA before the effective filing date of the claimed invention to combine Li et al. (‘649), Chen et al. (‘274), and Garimella et al. (‘625), as the combination teaches all elements of claim 6. Li et al. (‘649) teaches the method for detecting ghost objects in sensor measurements of a vehicle environment, generating object hypotheses having associated attributes, generating social context information from the relationships among the objects, searching for anomalies and patterns including comparisons of a first object’s attributes with the surrounding objects, and classifying objects as ghost objects. Chen et al. (‘274) teaches obtaining map context information from a digital road map, generating object hypotheses using several individual sensors, and comparing a radar detection from a further sensor with a camera-based object hypothesis to determine whether the detection is plausible or implausible. Garimella et al. (‘625) teaches storing map data, object attributes, and inter-object relational information as the nodes and edges of a graph structure, including the differences in attribute values between each object and the other objects. The motivations are drawn from the references’ own stated problems and advantages (Li et al. (‘649) [0005-0006]; Chen et al. (‘274) [0019]; Garimella et al. (‘625) [0011], [0022], [0029]). The combination uses each known element according to its established function to yield predictable results. See KSR, 550 U.S. at 416; MPEP § 2143(I)(A). A POSITA would have had a reasonable expectation of success in making this combination. Li et al. (‘649) demonstrates a successfully implemented ghost object detection method for automobile radar. Chen et al. (‘274) demonstrates a successfully implemented radar-camera verification system in an autonomous vehicle. Garimella et al. (‘625) demonstrates a successfully implemented graph structure representing combined map and entity data in an autonomous vehicle. Because these techniques operate on compatible data, namely object detections, their attributes, and their relative quantities, a POSITA would have reasonably expected the combination to produce a functional ghost object detection system.
Regarding Claim 7, Li et al. (‘649) in view of Chen et al. (‘274) and Garimella et al. (‘625) teaches the method of claim 6.
wherein the features of the graph structure include at least one of: geometric relationships between the plurality of objects, geometric relationships between the plurality of objects and locations of the plurality of objects in the environment, kinematic relationships between the plurality of objects that are dynamic objects, semantic relationships between the plurality of objects, or static relationships between the plurality of objects.
Claim 7 recites the features in the alternative (“at least one of … or”). Accordingly, it is sufficient that the art teaches one of the recited alternatives. The examiner addresses the first alternative, geometric relationships between the plurality of objects.
Li et al. (‘649) teaches geometric relationships between the plurality of objects: Li et al. (‘649) determines the colinearity of the objects with the radar ([0090]: “verifying, at the hardware processor 112, that two objects and radar are colinear by checking azimuth angle”) and the inter-object distances ([0091]: “determining, at the hardware processor 112, the distance from O1 to O2”).
Li et al. (‘649) does not explicitly teach these geometric relationships as features of a graph structure, but Garimella et al. (‘625) teaches geometric relationships between the objects as edge features of the graph structure: Garimella et al. (‘625) teaches edge data ([0021]: “including the relative distance, relative yaw, relative velocity, relative pose, relative size, relative acceleration, relative permissibility, and the like”). Because the “geometric relationships between the plurality of objects” alternative is taught by the combination of Li et al. (‘649) and Garimella et al. (‘625), claim 7 is met. The remaining alternatives need not be addressed.
It would have been obvious to a POSITA before the effective filing date of the claimed invention to store the geometric relationships computed by Li et al. (‘649) as edge features of the graph structure of Garimella et al. (‘625). The motivation and reasonable expectation of success are the same as those set forth above for claim 6, namely Garimella et al. (‘625)’s teaching that its edges natively store relative attribute differences between object pairs ([0022]), and its demonstrated implementation of such edge features in an autonomous vehicle system ([0021], [0045]).
Regarding Claim 8, Li et al. (‘649) in view of Chen et al. (‘274) and Garimella et al. (‘625) teaches the method of claim 6.
wherein the features of the graph structure include at least one of: temporal information, a spacing of the vehicle from at least one of the plurality of objects in the environment during a first detection of one of the plurality of objects or during first sensor measurements of a measurement series, a speed of at least one of the plurality of objects compared with further surrounding objects, a position of at least one of the plurality of objects compared with a road network determined from the map context information, or an orientation of at least one of the plurality of objects compared with an orientation of a lane located under the at least one of the plurality of objects.
Claim 8 recites the features in the alternative (“at least one of … or”). Accordingly, it is sufficient that the art teaches one of the recited alternatives. The examiner addresses the first alternative, temporal information.
Li et al. (‘649) teaches temporal information as an attribute of the detected objects: Li et al. (‘649) uses each object’s lifetime in its ghost classification ([0095]: “finding the dynamic object with less RCS or lifetime among the two dynamic objects”).
Li et al. (‘649) does not explicitly teach temporal information as a feature of a graph structure, but Garimella et al. (‘625) teaches temporal information as a feature of the graph structure: Garimella et al. (‘625) teaches that the graph holds ([0087]; see also [0026]: “the current state of the GNN at a first point in time (e.g., t=0)”, including “the attributes and/or features associated with the nodes and/or edges in the GNN structure”, and that the graph is updated when “the autonomous vehicle may perceive the entity at a different time, having a different position, velocity, or trajectory”). These time-indexed node and edge states constitute temporal information as a feature of the graph structure. Because the “temporal information” alternative is taught by the combination of Li et al. (‘649) and Garimella et al. (‘625), claim 8 is met. The remaining alternatives need not be addressed.
It would have been obvious to a POSITA before the effective filing date of the claimed invention to store Li et al. (‘649)’s lifetime attribute and the time-indexed object states as features of the graph structure of Garimella et al. (‘625). The motivation and reasonable expectation of success are the same as those set forth above for claim 6, and are further supported by Garimella et al. (‘625)’s demonstrated updating of node and edge states over successive observations ([0026], [0087]).
Regarding Claim 9, Li et al. (‘649) in view of Chen et al. (‘274) and Garimella et al. (‘625) teaches the method of claim 6.
Li et al. (‘649) does not explicitly teach wherein the classification is performed using learning by a graph-based artificial neural network, but Chen et al. (‘274) and Garimella et al. (‘625) together teach this limitation.
Chen et al. (‘274) teaches performing the classification of radar detections using learning: Chen et al. (‘274) teaches that RVM 132 ([0025]: “can include one or more trained models that are used to process some or all of the above data to classify the radar returns as genuine or spurious,” and that the models “can be trained using multiple annotated camera images and multiple sets of radar data, to identify specific camera and radar features in the respective images/data associated with genuine and false objects”).
Garimella et al. (‘625) teaches a graph-based artificial neural network operating on the graph structure: Garimella et al. (‘625) teaches that the GNN component ([0045]: “may include functionality to execute inference operations on the graph structure, using machine-learning techniques to determine inferred updated states of the nodes and edge features within the graph structure”). Garimella et al. (‘625) further teaches that ([0060]: “a neural network can utilize machine learning, which can refer to a broad class of such algorithms in which an output is generated based on learned parameters”).
It would have been obvious to a POSITA before the effective filing date of the claimed invention to perform the ghost object classification of the combined system using learning by a graph-based artificial neural network. Chen et al. (‘274) provides the motivation to perform the classification of genuine and false objects with trained models, because such models learn the ([0025]: “features … associated with genuine and false objects”). Garimella et al. (‘625) provides the motivation to perform the learned inference on the graph structure with a GNN, because ([0029]: “an execution of a single inference process (or operation) on the GNN may output an inferred updated representation of the environment as a whole”) with ([0011]: “technical advantages in reducing computing resources and improving efficiency”). These motivations are grounded in the references’ own teachings and do not rely on knowledge of the claimed invention. The reasonable expectation of success is established by Chen et al. (‘274)’s demonstrated trained classification of radar detections in an autonomous vehicle and Garimella et al. (‘625)’s demonstrated learned GNN inference over the same type of graph structure used in the combination.
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).
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 REMASH R GUYAH whose telephone number is (571)270-0115. The examiner can normally be reached M-F 7:30-4:30.
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, Resha H Desai can be reached at (571) 270-7792. 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.
/REMASH R GUYAH/Examiner, Art Unit 3648
/RESHA DESAI/Supervisory Patent Examiner, Art Unit 3648
