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 .
Detailed Action
This action is in response to the claims filed 10/21/2025:
Claims 1 – 14 are pending.
Claims 1 and 8 are independent.
Priority
Applicant’s claim for the benefit of a prior-filed application under 35 U.S.C. 119(e) or under 35 U.S.C. 120, 121, 365(c), or 386(c) is acknowledged. Applicant has not complied with one or more conditions for receiving the benefit of an earlier filing date under 35 U.S.C. 120, 121, 365(c), or 386(c) as follows:
The later-filed application must be an application for a patent for an invention which is also disclosed in the prior application (the parent or original nonprovisional application or provisional application). The disclosure of the invention in the parent application and in the later-filed application must be sufficient to comply with the requirements of 35 U.S.C. 112(a) or the first paragraph of pre-AIA 35 U.S.C. 112, except for the best mode requirement. See Transco Products, Inc. v. Performance Contracting, Inc., 38 F.3d 551, 32 USPQ2d 1077 (Fed. Cir. 1994).
The disclosure of the prior-filed application, Application No. 1905193, fails to provide adequate support or enablement in the manner provided by 35 U.S.C. 112(a) or pre-AIA 35 U.S.C. 112, first paragraph for one or more claims of this application. “maintain a latent manifold as a geometric substrate incorporating typed latent entities stratified according to structural properties, wherein local curvature reflects semantic density and entity types determine permissible operations;” and “execute type-aware geometric operations on the typed latent entities, wherein operation legality is determined by entity type and local manifold geometry, and wherein operations modify the manifold structure to reinforce successful cognitive patterns.” in claims 1 and 8. Similarly, the provisional 63/847,082 fails to provide adequate support or enablement in the manner provided by 35 U.S.C. 112(a) or pre-AIA 35 U.S.C. 112, first paragraph for one or more claims of this application. Therefore, the EFD is 10/21/2025.
Claim Rejections - 35 USC § 112
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 1-14 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 claims 1 and 8, "to reinforce successful cognitive patterns" is indefinite. "Successful" is a relative term of degree without a relative basis for comparison. In the interest of further examination all cognitive patterns are interpreted as successful cognitive patterns.
Regarding claims 7 and 14, "resist modification operations" is indefinite. "Resist" is a relative term of degree without any relative basis for comparison. In the interest of further examination all entities are interpreted as resisting modification operations.
Regarding claims 4 and 11, "high compressibility" is indefinite. "High" is a relative term of degree without any relative basis for comparison. In the interest of further examination any compressibility is interpreted as high compressibility.
The remaining claims are rejected with respect to their dependence on the rejected claims.
Claim Rejections - 35 USC § 101
101 Rejection
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 1-14 are rejected under 35 USC § 101 because the claimed invention is directed to non-statutory subject matter.
Regarding Claim 1: Claim 1 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 1 is directed to a computer system, which is directed to a product, one of the statutory categories.
Step 2A Prong One Analysis: Claim 1 under its broadest reasonable interpretation is a series of mental processes. For example, but for the generic computer components language, the above limitations in the context of this claim encompass machine learning processing, including the following:
maintain a latent manifold as a geometric substrate incorporating typed latent entities stratified according to structural properties, wherein local curvature reflects semantic density and entity types determine permissible operations; (observation, evaluation, and judgement),
implement temporal synchronization of heterogeneous multimodal data streams through generation of temporal alignment fields within the latent manifold that coordinate asynchronous inputs while preserving semantic coherence across modal boundaries (observation, evaluation, and judgement)
execute type-aware geometric operations on the typed latent entities, wherein operation legality is determined by entity type and local manifold geometry, and wherein operations modify the manifold structure to reinforce successful cognitive patterns (observation, evaluation, and judgement)
Therefore, claim 1 recites an abstract idea which is a judicial exception.
Step 2A Prong Two Analysis: Claim 1 recites additional elements “A computer system for persistent cognitive computation with temporally synchronized multimodal processing, comprising a hardware memory, wherein the computer system is configured to execute software instructions stored on nontransitory machine-readable storage media that”. However, these additional features are computer components recited at a high-level of generality, such that they amount to no more than mere instructions to apply the judicial exception using a generic computer component. An additional element that merely recites the words “apply it” (or an equivalent) with the judicial exception, or merely includes instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea, does not integrate the judicial exception into a practical application (See MPEP 2106.05(f)). Therefore, claim 1 is directed to a judicial exception.
Step 2B Analysis: Claim 1 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the lack of integration of the abstract idea into a practical application, the additional elements recited in claim 1 amount to no more than mere instructions to apply the judicial exception using a generic computer component.
For the reasons above, claim 1 is rejected as being directed to non-patentable subject matter under §101. This rejection applies equally to independent claim 8 which recites a method, as well as to dependent claims 2-7 and 9-14. The additional limitations of the dependent claims are addressed briefly below:
Dependent claims 2 and 9 recite additional observation, evaluation, and judgement “the temporal alignment fields are computed using differential geometry operations on Riemannian manifolds with variable curvature tensors”
Dependent claims 3 and 10 recite additional observation, evaluation, and judgement “the geometric operations include geodesic path computation that minimizes a cognitive action functional incorporating kinetic energy and compression pressure terms”
Dependent claims 4 and 11 recite additional observation, evaluation, and judgement “the typed latent entities comprise at least FACT entities characterized by atomic structure and high compressibility, TRAJECTORY entities comprising temporally ordered sequences with smooth continuity constraints, and AFFECT entities exhibiting field-like persistence with temporal decay properties”
Dependent claims 5 and 12 recite additional observation, evaluation, and judgement “the type-aware geometric operations include recombination operations that are permitted only when entities satisfy compatibility predicates based on geometric proximity and semantic alignment metrics”
Dependent claims 6 and 13 recite additional observation, evaluation, and judgement “the temporal synchronization generates compression pressure fields that vary continuously across the manifold based on local semantic density and modality-specific information characteristics”
Dependent claims 7 and 14 recite additional observation, evaluation, and judgement “the type-aware geometric operations are governed by an operational grammar that defines legal transformations for each entity type, wherein FACT entities permit generalization operations, TRAJECTORY entities permit splice operations only when endpoints satisfy continuity conditions, and ANCHOR entities resist modification operations”
Therefore, when considering the elements separately and in combination, they do not add significantly more to the inventive concept. Accordingly, claims 1-14 are rejected under 35 U.S.C. § 101.
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 for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-14 are rejected under U.S.C. §103 as being unpatentable over the combination of Soleimani (“Activity Recognition via Multimodal Large Language Models and Riemannian Optimization”, 2025) and Yan (US20250077872A1).
Regarding claim 1, Soleimani teaches A computer system for persistent cognitive computation with temporally synchronized multimodal processing, ([p. 1] "Abstract—Human Activity Recognition (HAR) has become a critical task in applications such as healthcare, smart environments, and human-computer interaction. This study investigates the potential of using a publicly available GPT-2 variant for HAR with multimodal data, combining its natural language processing capabilities with advanced temporal modeling for sequential data.")
maintain a latent manifold as a geometric substrate ([p. 5] "the modality-specific features are projected into a common embedding space of dimension dmodel and concatenated […] Treating the data and parameters as points on a Riemannian manifold enables the design of loss functions and optimization procedures that respect the underlying geometry" [p. 7] "the model parameters θ to lie on a Riemannian manifold M with metric tensor gθ [...] Expθt (·) maps a tangent vector back onto the manifold M")
incorporating typed latent entities stratified according to structural properties, ([p. 4] "To handle the diverse data types, specialized encoders are employed for each modality" [p. 5] "Skeletal data is represented as a graph G=(V,E)" [p. 5] "Inertial sensor readings are sequential data processed using a recurrent neural network (RNN), such as an LSTM" [p. 6] "We use a pre-trained LLM to encode text data" Soleimani explicitly creates typed internal representations)
wherein local curvature reflects semantic density ([p. 6] "The resulting textual embeddings are incorporated into the multimodal feature fusion process." [p. 7] "Riemannian regularization offers several key advantages for multimodal HAR tasks: (i) It aligns with the intrinsic geometric structure of multimodal data, such as videos, skeletons, and sensor signals, by optimizing parameters on curved manifolds" Soleimani explicitly uses a LLM embeddings encoding activity descriptions and contextual information (semantic content), where related activities from nearby or well-defined clusters are determined (semantic density) (See Table I and FIG. 2-3) which are learned through Riemannian regularization having local curvature)
and entity types determine permissible operations;([p. 4] "Each modality is processed through a modality-specific encoder to extract features suitable for fusion")
implement temporal synchronization of heterogeneous multimodal data streams ([p. 1] "GPT-2 variant for HAR with multimodal data, combining its natural language processing capabilities with advanced temporal modeling for sequential data." [p. 4] "Temporal alignment is essential when dealing with multimodal data collected asynchronously. To achieve this, data streams are synchronized by resampling each modality to a common frame rate fs, ensuring that corresponding time steps across modalities accurately represent the same moment in the action sequence.")
through generation of temporal alignment fields within the latent manifold that coordinate asynchronous inputs ([p. 4] "Temporal alignment is essential when dealing with multimodal data collected asynchronously [...] ensuring that corresponding time steps across modalities accurately represent the same moment" [p. 5] "The modality-specific features are projected into a common embedding space of dimension d model and concatenated: […] To incorporate temporal information, positional encodings are added to the embeddings […] pt∈R4dmodel is a learnable vector associated with time step t." Eqn. 12 creates latent multimodal embeddings, Eqn. 13 then generates or learns a time-indexed family and applies it to that latent embedding to create zt. A mathematical field assigns a value to each point in an index domain. Pt is interpreted as a discrete temporal alignment field within the latent representation)
while preserving semantic coherence across modal boundaries; and ([p. 4] "Temporal alignment is essential when dealing with multimodal data collected asynchronously [...] ensuring that corresponding time steps across modalities accurately represent the same moment" [p. 7] "The resulting textual embeddings are incorporated into the multimodal feature fusion process")
execute type-aware geometric operations on the typed latent entities, ([p. 5] "specialized encoders are employed for each modality" [p. 7] "We adopt Riemannian stochastic gradient descent (RSGD) to update the model parameters" Eqn. 27 and 28 performed in Algorithm 1 interpreted as geometric operations which update the model containing modality-specific encoders, projections, temporal vectors, and fused latent representations. This is interpreted as a type-aware geometric computation)
wherein operation legality is determined by entity type and local manifold geometry, ([p. 7] "PTθM is the projection operator onto the tangent space. The parameters are updated using the Riemannian exponential map […] where Expθt (·) maps a tangent vector back onto the manifold M" [p. 10] "dynamic padding
was applied within each batch to ensure compatibility with modality-specific encoders")
and wherein operations modify the manifold structure to reinforce successful cognitive patterns. ([p. 8] "The optimization algorithm updates the parameters along geodesics on the manifold using the Riemannian exponential map Expθ").
However, Soleimani does not explicitly teach comprising a hardware memory, wherein the computer system is configured to execute software instructions stored on nontransitory machine-readable storage media that:.
Yan, in the same field of endeavor, teaches comprising a hardware memory, wherein the computer system is configured to execute software instructions stored on nontransitory machine-readable storage media that:([¶0021] "a signal detection system is provided that includes one or more memory storage devices, and a processor-based device in electrical communication with the one or more memory storage devices.").
Soleimani as well as Yan are directed towards using Riemannian gradient descent to optimize machine learning models. Therefore, Soleimani as well as Yan are analogous art in the same field of endeavor. It would have been obvious before the effective filing date of the claimed invention to combine the teachings of Soleimani with the teachings of Yan by applying the system in Soleimani on the computer hardware described by Yan. Yan provides as additional motivation for combination ([¶0090] “The controller device is configured to facilitate, for example, signal detection using an optimized template determined with a trainable machine learning system. The storage device may thus include a computer program product that when executed on the controller device (which, as noted, may be a processor-based device) causes the processor-based device to perform operations to facilitate the implementation of procedures and operations described herein”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 2, the combination of Soleimani, and Yan teaches The computer system of claim 1, wherein the temporal alignment fields are computed using differential geometry operations (Soleimani [p. 5] "learnable positional embeddings are adopted" [p. 6] "We adopt Riemannian stochastic gradient descent (RSGD) to update the model parameters" [p. 8] "unlike L1, which is non differentiable at zero, L2 is differentiable everywhere, simplifying implementation and avoiding potential optimization instabilities. In summary, the combination of Riemannian regularization and the L2 norm ensures robust, efficient, and generalizable learning, making it an ideal choice for addressing the complexities of multimodal HAR tasks")
on Riemannian manifolds with variable curvature tensors.(Soleimani [p. 6] "the model parameters θ to lie on a Riemannian manifold M with metric tensor gθ" [p. 7] "n representing the normal vector to the manifold at θ, and ⟨·,·⟩ denoting the inner product induced by the metric gθ […] M is the Stiefel manifold (the set of orthonormal matrices)" Soleimani defines its geometry by the local metric tensor, indexed by the current manifold point/model state theta. That metric governs the tangent projection, inner product, gradient norm, geodesic, and exponential-map update. Riemannian curvature is mathematically induced by the local metric and its variation over the manifold.).
Regarding claim 3, the combination of Soleimani, and Yan teaches The computer system of claim 1, wherein the geometric operations include geodesic path computation that minimizes a cognitive action functional incorporating kinetic energy (Soleimani [p. 6] "LRM = 1/2 ∥∇MLCE∥2 gθ" LRM is in the standard mathematical form of kinetic energy)
and compression pressure terms.(Soleimani [p. 7] "λLRM […] The L2 norm provides smooth gradients and gradual updates, ensuring stable and efficient optimization, particularly for large models like Transformers. It prevents any single modality from dominating the learning process" Functionally λLRM (interpreted as compression pressure term) exerts optimization pressure that constrains or compresses the parameter configuration while balancing the modalities.).
Regarding claim 4, the combination of Soleimani, and Yan teaches The computer system of claim 1, wherein the typed latent entities comprise at least FACT entities characterized by atomic structure and high compressibility, (Soleimani [p. 6] "We use a pre-trained LLM to encode text data […] The text may include descriptions of the activities or contextual information" [p. 7] "The L2 norm provides smooth gradients and gradual up dates, ensuring stable and efficient optimization, particularly for large models like Transformers. It prevents any single modality from dominating the learning process" textual activity descriptions, class labels, and LLM embeddings interpreted as FACT entities)
TRAJECTORY entities comprising temporally ordered sequences with smooth continuity constraints, (Soleimani [p. 5] "zt=et+pt […] where pt is a learnable vector associated with time step t […] The sequence of embeddings {z1,z2,…,zT} is processed through L Transformer encoder layers" [p. 7] "The optimization algorithm updates the parameters along geodesics on the manifold using the Riemannian exponential map Expθ" The Riemannian updates follow smooth geodesics)
and AFFECT entities exhibiting field-like persistence with temporal decay properties.(Soleimani [p. 4] "Inertial sensor readings Rt are filtered using a low-pass filter to remove high-frequency noise" The preceding state Rt-1 carries forward (persistence) with decay properties (1-a). The indexed family t->Rt is interpreted as a discrete time-dependent field. For these reasons the inertial/LSTM state Rt is interpreted as the AFFECT entity.).
Regarding claim 5, the combination of Soleimani, and Yan teaches The computer system of claim 1, wherein the type-aware geometric operations include recombination operations that are permitted only when entities satisfy compatibility predicates based on geometric proximity and semantic alignment metrics.(Soleimani [p. 5] "the modality-specific features are projected into a common embedding space of dimension dmodel and concatenated" [p. 7] "The textual features are projected into the common embedding space and concatenated with other modality features" Soleimani recombines modality-specific entities after applying type-specific projections. The attention head in Eqn. 15 comprises a geometric similarity/proximity score in the learned embedding space which Softmax converts into the degree of contribution).
Regarding claim 6, the combination of Soleimani, and Yan teaches The computer system of claim 1, wherein the temporal synchronization generates compression pressure fields that vary continuously across the manifold based on local semantic density and modality-specific information characteristics.(Soleimani [p. 4] "data streams are synchronized by resampling each modality to a common frame rate fs, ensuring that corresponding time steps across modalities accurately represent the same moment in the action sequence" Soleimani discloses this through synchronized inputs producing a smooth Riemannian regularization field shaped by semantic and modality-specific features. More specifically, because LRM assigns a regularization value to each model state it functions as a compression field over the manifold. Its "pressure" function is supported by Soleimani's explanation that "It prevents any single modality from dominating the learning process" ([p. 7])).
Regarding claim 7, the combination of Soleimani, and Yan teaches The computer system of claim 1, wherein the type-aware geometric operations are governed by an operational grammar that defines legal transformations for each entity type, (Soleimani [p. 7] "PTθM is the projection operator onto the tangent space. The parameters are updated using the Riemannian exponential map […] where Expθt (·) maps a tangent vector back onto the manifold M" [p. 10] "dynamic padding
was applied within each batch to ensure compatibility with modality-specific encoders")
wherein FACT entities permit generalization operations, (Soleimani [p. 6] "We use a pre-trained LLM to encode text data […] The text may include descriptions of the activities or contextual information" [p. 7] "The L2 norm provides smooth gradients and gradual up dates, ensuring stable and efficient optimization, particularly for large models like Transformers. It prevents any single modality from dominating the learning process" textual activity descriptions, class labels, and LLM embeddings interpreted as FACT entities)
TRAJECTORY entities permit splice operations only when endpoints satisfy continuity conditions, (Soleimani [p. 5] "zt=et+pt […] where pt is a learnable vector associated with time step t […] The sequence of embeddings {z1,z2,…,zT} is processed through L Transformer encoder layers" [p. 7] "The optimization algorithm updates the parameters along geodesics on the manifold using the Riemannian exponential map Expθ" The Riemannian updates follow smooth geodesics)
and ANCHOR entities resist modification operations.(Soleimani [p. 7] "λLRM […] The L2 norm provides smooth gradients and gradual updates, ensuring stable and efficient optimization, particularly for large models like Transformers. It prevents any single modality from dominating the learning process").
Regarding claims 8-14, claims 8-14 are directed towards the method performed by the system of claims 1-7, respectively. Therefore, the rejections applied to claims 1-7 also apply to claims 8-14.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Ghosh (“The Geometry of Understanding: A Philosophical Blueprint for Provably Faithful Neuro-Symbolic Translation”, 2025) is directed towards using Riemannian optimization for large language models.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SIDNEY VINCENT BOSTWICK whose telephone number is (571)272-4720. The examiner can normally be reached M-F 7:30am-5:00pm EST.
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, Miranda Huang can be reached on (571)270-7092. 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.
/SIDNEY VINCENT BOSTWICK/Examiner, Art Unit 2124
Prosecution Insights