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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on February 23, 2026 has been entered.
Response to Amendment
Receipt is acknowledged of claim amendments with associated arguments/remarks, received February 23, 2026. Claims 1-7, 9-10 are pending with amendments to claim 1 and introduce new claim 10. Claim 8 was cancelled.
Response to Arguments
Applicant’s arguments, see Remarks, pg 5, filed 02/23/2026, with respect to the rejection of claim 8 under 35 USC § 112(a) has been fully considered but is moot because the claim is cancelled. Therefore, the rejection has been withdrawn.
Applicant’s arguments, see pg 5-9, filed 02/23/2026, with respect to the rejections of claim 1-4, 7-9 under 35 USC § 103 have been fully considered but are not persuasive.
Regarding Claim 1, the applicant first argues the cited art does not teach “learning a sparse coding dictionary by the computer on an input signal (14) to obtain a representation of low-complexity components within the invariant representations” and cites to the specification pg 2 ln 21 – pg 3 ln 17 to support of the claim amendment (Remarks – 02/23/2026 pg 5 ln 22 – pg 6 ln 19).
The applicant first argues “the definition of low-complexity components should be understood in the context of the entire specification” (Remarks – 02/23/2026 pg 6 ln 10-11) and “the specification sets forth clearly what the term “low-complexity components” in the context of a “hierarchically structured entity” means” (Remarks – 02/23/2026 pg 6 ln 20-21).
The specification states “Hierarchical means in that sense, for example: pixels form lines, lines form polygons, polygons form 3D bodies, and so on” (Specification, pg 3 ln 2-3) and that a “signal represents an hierarchically structured entity, e.g. in the case of images a hierarchy of “could be pixels” – lines – polygons – 3D-objects.” (Specification, pg 3 ln 9-11). The specification further recites that “if the 3D object is a cube, from the desired representation it is obvious that there is a cube, independent of the position of the cube” (Specification, pg 3 ln 15-17).
The definition of what a “low-complexity component” is therefore unclear and not defined in a meaningful way for one of ordinary skill in the art to readily know the scope and limitations of interpreting “low-complexity”. The claim, based on the specification, is unclear if “low-complexity” limitation is meant as a pixel, a line, a polygon, a 3D body, a cube and so on. Does low-complexity mean any level on a hierarchy of components or is there a point when a “low-complexity” becomes a “mid-complexity” or “high-complexity” component? The claim does not define a meaningful limit. The specification does not define the limit of the relative term “low-complexity.”
Accordingly, under MPEP § 2173.05(b), “When relative terms are used in claims wherein the improvement over the prior art rests entirely upon size or weight of an element in a combination of elements, the adequacy of the disclosure of a standard is of greater criticality.”
In this case, the claim does not clearly define the scope for the term “low-complexity components” and the specification does not disclose a meaningful standard that is apparent to one of ordinary skill in the art. Therefore, the claim is rejected for indefiniteness.
Regarding the applicant’s argument the cited prior art, Paiton et al (US 2014/0072213) does not teach “to obtain a representation of low-complexity components within the invariant representations”, the examiner respectfully disagrees.
The office action (Final Rejection – 12/03/2025) citation is not to the argued paragraphs cited by the applicant. The office action cites Figs 1, 2 and ¶ [0045]-[0052], which describes “the first layer of dictionary elements may be trained on small image patches, whereas subsequent layers may be trained on a combination of direct feature-based inputs along with down-sampled, pixel-based inputs arising from the images generated by the previously layer's sparsely activated features” (¶ [0047]). The image data is thus identified and stored in “hierarchically-organized learned dictionaries” with the higher layers “direct feature-based inputs’ and the lowest-layers refer to “pixel-based inputs” (¶ [0046]-[0047]). The hierarchical analysis is performed such that invariant elements are also detected (“embodiments may subdivide the layers into two stages, including a first stage of selective feature detectors and a second stage of invariant elements that may be implemented by combining local and lateral dictionaries” ¶ [0042]; further described in ¶ [0054]-[0059] and Fig 4, 5).
Respectfully, the examiner is not persuaded by the applicant that the prior art does not teach “ to obtain a representation of low-complexity components within the invariant representations.”
Applicant next argues the cited prior art does not teach “hierarchical” structured entities as claimed by applicant and argues the hierarchical approach of the prior art is with regard to the structure of the algorithms, not with the process of data analysis (Remarks – pg 6 ln 20 – pg 8 ln 12).
The examiner notes the limitation argued (“hierarchically structured entities”) is not in fact contained within the limitation in question (“learning a sparse coding dictionary by the computer on an input signal (14) to obtain a representation of low-complexity components within the invariant representations”). As noted by the applicant (Remarks – pg 8 ln 6-7), the preamble states the image processing is applied to “invariant representations of hierarchically structured entities.”
Regarding applicant’s argument that “Paiton's use of the term "hierarchical" refers to the model layers and not to any 2D/3D objects in the processed images recited in claim 1” (Remarks – pg 7 ln 19-20), the examiner is not persuaded.
The applicant’s arguments focus on use of citations to the background in the prior art (applicant cites prior art ¶ [0003]-[0006], [0009], [0011] in Remarks pg 7 ln 5 – pg 8 ln 5). The argument focuses on complexity of hierarchical networks without considering the context of the hierarchical layers in the networks “in order to capture the increased dimensionality and complexity of their inputs” (Paiton et al ¶ [0006]), which does appear to be the same concept as applicant’s claim language for hierarchically structured entities – that dimensionality increases from the inputs with continued analysis (as applicant’s specification describes the hierarchy as a pixel, a line, a polygon, Specification pg 3 ln 2-3).
Furthermore, the office action (Final Rejection – 12/03/2025) citation is not to the argued paragraphs cited by the applicant. The office action cites Figs 1, 2 and ¶ [0045]-[0052], which describes a pixel-based and a feature-based dictionary used for hierarchical processing of image data with the hierarchically-organized learned dictionaries of the image data based on visual features with layers of complexity allow for “representations that are approximately three times more sparse, while supporting equivalent or slightly superior image reconstruction quality” (¶ [0046]). It is understood these dictionaries are then used for hierarchical models in analyzing the image data, thereby “allowing higher layers to encode lower spatial-frequency visual features while lower layers encode higher spatial-frequency details” (¶ [0046]). The hierarchical analysis is performed such that invariant elements are also detected (“embodiments may subdivide the layers into two stages, including a first stage of selective feature detectors and a second stage of invariant elements that may be implemented by combining local and lateral dictionaries” ¶ [0042]; further described in ¶ [0054]-[0059] and Fig 4, 5).
The examiner did not readily identify in the applicant’s specification how the “sparse coding dictionary” claimed is structured or organized different from the prior art. Applicant’s specification describes that layers are used in a neural network (layer 1 (7), layer 2 (11)) and a hierarchical process such that the first layer analyzes an overall input image and the second layer narrows in in the pattern of the image to identify “pixel lines” (Specification pg 11 ln 24 – pg 12 ln 20).
The claim limitation of “hierarchically structured entities” can be broadly interpreted as any level of data that may be organized as the applicant has argued visual data to range from pixel to entire objects with invariant representation, and as discussed above, there is no meaningful limits to this terminology. Furthermore, the applicant states the broad sweeps of interpretation for the concept (Specification pg 12 ln 22-27):
In principle, possible further preferred embodiments could comprise of very different software products which use the described method, for example, to perform tasks like image denoising, object recognition, speech recognition, etc. The most immediate examples could be methods and constitutive systems which perform special cases of text recognition, for example, to solve Captchas or to recognize chemical structures in images.
Respectfully, the examiner is not persuaded that the prior art does not teach “hierarchically structured entities” in parallel to the applicant’s claimed invention.
Moreover, after further consideration of the claim language, the specification and applicant’s arguments, the term “hierarchically structured entities” is rejected for indefiniteness as it is unclear if the applicant claims this terminology in regard to the algorithm structure or in regard to the organization of the data contained in the dictionary and is rejected for indefiniteness.
Applicant further argues the cited prior art does not teach "inferring possible transformations from statistics of the sparse representation by computing a correlation matrix (8) between the low-complexity components with the computer resulting in invariance transformation of the data now encoded in symmetries of the correlation matrix (8)." (Remarks – pg 8 ln 13 – pg 9 ln 14).
The examiner notes “the sparse representation” lacks antecedent basis because the prior limitation recites “a sparse coding dictionary by the computer on an input signal (14) to obtain a representation of low-complexity components within the invariant representations”. Is “sparse representation” referring to the “sparse coding dictionary” or “invariant representations” or the “low-complexity components”? After careful consideration, the claim limitation “the sparse representation” lacks antecedent basis and is rejected for indefiniteness.
The examiner also notes the claim limitation recites “invariance transformation of the data” is “now encoded.” However, no step was recited in the claim to encode data and it is unclear as to what data is encoded – is it referring to the “sparse coding dictionary” or “invariant representations” or the “low-complexity components”? After careful consideration, the claim limitation “the data now encoded” lacks antecedent basis and is rejected for indefiniteness.
The applicant argues the “correlation matrix” is based on a “count” for a given pair of two neurons to be activated together (Remarks pg 8 ln 18 – pg 9 ln 7). However, the claim language does not reflect the features in which the applicant argues.
In response to applicant's argument that the references fail to show certain features of the invention, it is noted that the features upon which applicant relies (i.e., the correlation matrix 'count' how often a given pair of atoms is activated simultaneously by the same input data point) are not recited in the rejected claim(s). Although the claims are interpreted in light of the specification, limitations from the specification are not read into the claims. See In re Van Geuns, 988 F.2d 1181, 26 USPQ2d 1057 (Fed. Cir. 1993).
The prior art Paiton describes the hierarchically-organized learned dictionaries are comprised of “representations that are more invariant to small image transformations” (¶ [0042]) and, by generating the dictionaries in a top-down competition, the representations are “sparse” (¶ [0046], shown in Fig 2 element 200 and ¶ [0051]). The transformations of the representations are inferred between the layers (V1 and V2, V2 and V4; Fig 1) by calculating the difference between the hierarchical layer (error) using a respective matrix B2 and B4 (¶ [0049]-[0050]). This is further described such that local image transformations are more invariant from a local and lateral connection when the error layer is zero (Fig 4 and ¶ [0058]) and understood equivalent that when the error between the layers is zero there is an equivalent “symmetries of the correlation matrix” of sparse representation with invariance transformation.
If the applicant intends to have a specific definition intended for the interpretation of a limitation, such as the applicant’s argument that the “correlation matrix” is based on a “count” for a given pair of two neurons to be activated together, the examiner suggests clarifying the claim language to include such definition.
Respectfully, the examiner is not persuaded.
Applicant’s arguments of secondary references are dependent on arguments pertaining to Paiton, and are not persuasive as described above. No specific argument was raised regarding a secondary reference teaching.
No additional arguments are presented.
All arguments were addressed.
Claim Rejections - 35 USC § 112(b)
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 1 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.
Claim 1 recites the limitation “low-complexity components” in the limitation “learning a sparse coding dictionary by the computer on an input signal (14) to obtain a representation of low-complexity components within the invariant representations.” As described above in the remarks it is unclear as to the scope of the relative term “low-complexity.” The claim, based on the specification, is unclear if “low-complexity” limitation is meant as a pixel, a line, a polygon, a 3D body, a cube and so on. Does low-complexity mean any level on a hierarchy of components or is there a point when a “low-complexity” becomes a “mid-complexity” or “high-complexity” component? The claim does not define a meaningful limit. The specification does not define the limit of the relative term “low-complexity.”
Accordingly, under MPEP § 2173.05(b), “When relative terms are used in claims wherein the improvement over the prior art rests entirely upon size or weight of an element in a combination of elements, the adequacy of the disclosure of a standard is of greater criticality.” In this case, the claim does not clearly define the scope for the term “low-complexity components” and the specification does not disclose a meaningful standard that is apparent to one of ordinary skill in the art. Thus, the Applicant has failed to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor regards as the invention. Therefore, the claim is rejected for indefiniteness.
Claim 1 recites the limitation “hierarchically structured entities” in the preamble “A method for processing digital image recognition of invariant representations of hierarchically structured entities” and last step limitation “using the trained artificial neural network to the digital image recognition of hierarchically structured entities.” After consideration of applicant’s Remarks and review of the specification, is it unclear if “hierarchically structured entities” is with regard to the system that processes the image data (described with specific layers and steps, thereby hierarchically structured, shown in Fig 3) or with regard to the processed data (described with discrete patterns with representations ranging from pixel to object). Because the limitation “hierarchically structured entities” could be interpreted as either the system or the data, the limitation is unclear. Thus, the Applicant has failed to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor regards as the invention. Therefore, the claim is rejected for indefiniteness.
Claim 1 recites the limitation “repeating with step one” in the limitation “repeating with step one with the next higher hierarchy level (11) until all hierarchy levels (7, 11) of the invariant representations of the hierarchically structured entities are processed and the artificial neural network is trained.” However, “step one” was not clearly identified in the claim language and it is thus unclear which former step is meant as “step one.” Further, it is unclear if all “steps” are to be repeated or if only “step one” is repeated. Thus, the Applicant has failed to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor regards as the invention. Therefore, the claim is rejected for indefiniteness.
Claim 1 recites the limitation “the sparse representation” in the limitation "inferring possible transformations from statistics of the sparse representation by computing a correlation matrix (8) between the low-complexity components with the computer resulting in invariance transformation of the data now encoded in symmetries of the correlation matrix (8)." However, “the sparse representation” lacks antecedent basis because the prior limitation recites “a sparse coding dictionary by the computer on an input signal (14) to obtain a representation of low-complexity components within the invariant representations”. Is “sparse representation” referring to the “sparse coding dictionary” or “invariant representations” or the “low-complexity components”? After careful consideration, the claim limitation “the sparse representation” lacks antecedent basis. Thus, the Applicant has failed to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor regards as the invention. Therefore, the claim is rejected for indefiniteness.
Claims 2-7, 9-10 are rejected based on their dependency to claim 1.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-4, 7, 9-10 are rejected under 35 U.S.C. 103 as being unpatentable over Paiton et al (US 2014/0072213) in view of Bronstein et al (Geometric deep learning beyond Euclidean data, disclosed IDS 06/01/2023).
Regarding Claim 1, Paiton et al teach a method for processing digital image recognition of invariant representations of hierarchically structured entities, performed by a computer using an artificial neural network, comprising the following method steps:
- learning a sparse coding dictionary by the computer on an input signal (14) to obtain a representation of low-complexity components within the invariant representations (sparse representations of objects in undercomplete feature-based dictionaries are learned in a top-down hierarchy, and the dictionaries 200 are trained in a top-down methodology with a pixel-based dictionary; Figs 1, 2 and ¶ [0045]-[0052]; see Remarks above for further discussion),
- inferring possible transformations from statistics (interpreted as data representation of the image, not actual statistics; specification pg 4 ln 19) of the sparse representation by computing a correlation matrix (8) between the low-complexity components with the computer resulting in invariance transformation of the data now encoded in symmetries of the correlation matrix (8) (image reconstructions generated by sparse V1 and V2 activity and reconstructions generated by sparse V2 and V4 activity are compared using connectivity matrices B2 and B4, respectively to determine the difference, as error E2 and E4 (which may also be performed for small invariant image transformations ¶ [0055]-[0056]); Figs 1, 2 and ¶ [0049]-[0050]; see Remarks above for further discussion),
- repeating with step one with the next higher hierarchy level (11) until all hierarchy levels (7, 11) of the invariant representations of the hierarchically structured entities are processed and the artificial neural network is trained (the neurons of the network of the next layer down is trained for sparser, more invariant representations in the dictionary; Figs 1, 2, 4 and ¶ [0045]-[0052], [0058]), and
- using the trained artificial neural network to the digital image recognition of hierarchically structured entities, creating representations of those entities which are invariant under the transformations learnt in the previous steps (using a top-down hierarchy of representations, the dictionaries are learned for sparser, more invariant representations in training the neurons of the network with down-sampled, pixel-based inputs generated by the previous layer’s sparsely activated features; Figs 1, 2, 4, 5 and ¶ [0045]-[0052], [0058]-[0059]).
Paiton et al does not teach - computing eigenvectors (9) of the Laplacian operator on a graph (18) whose adjacency matrix is the correlation matrix (8) from the previous step, and -performing a coordinate transformation to a base of eigenvectors (9) of the Laplacian operator.
Bronstein et al is analogous art pertinent to the technological problem addressed in this application and teaches – computing eigenvectors (9) of the Laplacian operator on the graph (18) whose adjacency matrix is the correlation matrix (8) (no guidance was readily identified for “adjacency matrix”, specification pg 4, ln 24) from the previous step (a Laplacian eigenfunction optimization problem is applied to determine the eigenvectors, based on the domain, with k eigenvectors represented as
Φ
k
(Fig 3 and [IN3] Physical interpretation of Laplacian eigenfunctions(pg 9)), and for a CNN a Laplacian operator is applied on eigenfunctions of the point on the graph (16s16 image patch) using a correlation kernel (matrix); Fig 5a and [IN5] Rediscovering standard CNNs using correlation kernels (pg 12)), and
- performing a coordinate transformation to a base of eigenvectors (9) of the Laplacian operator (a coordinate transformation is performed based on the eigenvectors of the resulting graph Laplacian
Φ
Γ
l
,
l
'
Φ
T
, diagonalized by a standard DCT, resulting in translation invariant and classical convolutions of the image data; Fig 5a and [IN5] Rediscovering standard CNNs using correlation kernels (pg 12)),
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to combine the teachings of Paiton et al with Bronstein et al including -computing eigenvectors (9) of the Laplacian operator on the graph (18) whose adjacency matrix is the correlation matrix (8) from the previous step, and - performing a coordinate transformation to a base of eigenvectors (9) of the Laplacian operator. By determining the eigenvectors of a Laplacian operator and performing a coordinate transformation, the spatial arrangement of the pixels is recovered and enables the model to learn a number of parameters independent of the input size, thereby improving the model by requiring less parameters to recover meaningful image data, as recognized by Bronstein et al (Spectral CNN with smooth spectral multipliers ¶ 1).
Regarding Claim 2, Paiton et al in view of Bronstein et al teach the method according to claim 1 (as discussed above), wherein the sparse coding dictionary learning comprises a first processing step of recognizing patterns (15) in the input signal data (14) (Paiton et al, training of the undercomplete feature-based dictionary include learning long-range lateral pooling patterns , wherein those patterns (15) represent specific recurring combinations in the input signal data (14) (Paiton et al, sparse image-patch representations are based on sparse patterns, such as color or texture, with the dictionary elements learned based on the pattern; ¶ [0080]-[0082]).
Regarding Claim 3, Paiton et al in view of Bronstein et al teach the method according to claim 1 (as discussed above), wherein the representation of low-complexity components is created by computing a correlation matrix (8) of co-occurrences of neuron activations (no guidance was readily identified for “co-occurrences”, specification pg 5, ln 17-19) (Paiton et al, each level of the sparser representations are generated by the top-down feedback based on the degree of sparsity given by the fraction of neurons with non-zero activation, with sparse approximation generated used to adjust the connectivity matrices (also see ¶ [0076]-[0078], which discloses activation of features based on lateral connectivity and interactions based on co-occurrences of edges); ¶ [0050], [0053]).
Regarding Claim 4, Paiton et al in view of Bronstein et al teach the method according to claim 1 (as discussed above), wherein the next higher hierarchy level (11) gets the result of the coordinate transformation from the base of eigenvectors (9) as input data (Bronstein et al, the global and local regularity of the graph is combined to produce layers with constant number of parameters (thereby layer parameters does not depend on size) in classical Euclidean CNNs, thus the eigenvector basis
Φ
is consistent; Fig 2 and Spectral CNNs ¶ 3-5).
Regarding Claim 7, Paiton et al in view of Bronstein et al teach a computer configured to establish an artificial neural network by performing the method (Paiton et al, a computing system 1600 executes and stores computer-readable instructions, including a generative model used with a hierarchical dictionary-based network to perform object detection and tracking in image data; ¶ [0042]-[0046], [0062]) according to claim 1 (as described above).
Regarding Claim 9, Paiton et al in view of Bronstein et al teach the method according to claim 1 (as discussed above), wherein learning the sparse coding dictionary includes obtaining the representation of the low-complexity components using patterns of data in the input signal (Paiton et al, sparse representations of objects in undercomplete feature-based dictionaries are learned in a top-down hierarchy, and the feature-based dictionaries are processed based on patterns that can be created by combining different features across space; Figs 1, 2 and ¶ [0013], [0045]-[0052], [0056])
Regarding Claim 10, Paiton et al in view of Bronstein et al teach the method according to claim 1 (as discussed above), wherein a low complexity component of the low complexity components within the invariant representations is a line segment of a hierarchically structured entity (Paiton et al, image data of objects may be evaluated for shape/contour detection with the dictionary identification of lines for edge representation, with the detector and dictionary based on eight orientations between 0 and 180 degrees; Fig 7, 9 and ¶ [0067], [0076]-[0078]).
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Paiton et al (US 2014/0072213, previously cited in Final Rejection – 12/03/2025) in view of Bronstein et al (Geometric deep learning beyond Euclidean data, previously cited in Final Rejection – 12/03/2025) and Devries (US 2020/0401212, previously cited in Final Rejection – 12/03/2025).
Regarding Claim 5, Paiton et al in view of Bronstein et al teach the method according to claim 1 (as discussed above).
Paiton et al in view of Bronstein et al do not teach wherein the using of the trained artificial neural network to digital image recognition comprises image denoising, object recognition, speech recognition, and text recognition.
Devries is analogous art pertinent to the technological problem addressed in this application and teaches the using of the trained artificial neural network to digital image recognition comprises image denoising, object recognition, speech recognition, and text recognition (interpreted that any one of the steps (denoising, or recognition) is “using” the ANN) (digital image analysis is performed using a trained artificial neural network with sparse dictionary learning for object recognition and includes the digital image analysis and modifications based on typed input and/or speech input, and enhancement includes noise reduction (¶ [0043]-[0044], [0049], [0082]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to combine the teachings of Paiton et al in view of Bronstein et al with Devries including wherein the using of the trained artificial neural network to digital image recognition comprises image denoising, object recognition, speech recognition, and text recognition. By performing numerous operations using the neural network for performing image analysis, the algorithm is capable to effectively perform object recognition used in an augmented-reality system in which a user may readily interact, as recognized by Devries (¶ [0003]-[0004]).
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Paiton et al (US 2014/0072213, previously cited in Final Rejection – 12/03/2025) in view of Bronstein et al (Geometric deep learning beyond Euclidean data, previously cited in Final Rejection – 12/03/2025), Devries (US 2020/0401212, previously cited in Final Rejection – 12/03/2025) and Garg et al (Neural Network CAPTCHA Crackers, previously cited in Final Rejection – 12/03/2025).
Regarding Claim 6, Paiton et al in view of Bronstein et al and Devries teach the method according to claim 5 (as discussed above).
Paiton et al in view of Bronstein et al and Devries do not teach wherein, the text and object recognition comprises to solve captchas or to recognize chemical structures in images.
Garg et al is analogous art pertinent to the technological problem addressed in this application and teaches wherein, the text and object recognition comprises to solve captchas or to recognize chemical structures in images (the neural network with image object and text recognition is applied to solving CAPTCHAs; Figs 3-4 and V. Dataset, VI. Models).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to combine the teachings of Paiton et al in view of Bronstein et al and Devries with Garg et al including wherein, the text and object recognition comprises to solve captchas or to recognize chemical structures in images. By applying a trained artificial neural network with sparse layers (not fully connected) and applied to CAPTCHAs, the security of CAPTCHAs can be challenged to determine how to improve the systems, as recognized by Garg et al (I. Introduction).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Linde et al (A White Paper on the Future of Artificial Intelligence) describes, by the author in a publication released over a year before the earliest filing date of the current application, the use of invariant representations applied to neural networks as an unsupervised learning approach based on a six step algorithm to extrapolate observations for unsupervised learning of invariant representations as similarly claimed in the current application.
Poggio et al (US 2015/0278635, previously cited in Final Rejection – 12/03/2025) teach a system and method for processing image data including detecting a pattern based on stored pattern representations, including transformation, translation and scaling analysis.
Mehr et al (US 2018/0314917, previously cited in Final Rejection – 12/03/2025) teaches an encoding for image data including analysis of objects with transformations of object representations.
Spratling (A Hierarchical Predictive Coding Model of Object Recognition in Natural Images, previously cited in Final Rejection – 12/03/2025) teach a hierarchical neural network based on predictive coding for performing visual object recognition.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KATHLEEN M BROUGHTON whose telephone number is (571)270-7380. The examiner can normally be reached Monday-Friday 8:00-5:00.
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, John Villecco can be reached at (571) 272-7319. 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.
/KATHLEEN M BROUGHTON/Primary Examiner, Art Unit 2661
Prosecution Insights