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 .
Claims 1-10 are presented for examination.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on October 19, 2023 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Drawings
The drawings are objected to because Figure 9 contains text written on a shaded background, see 37 CFR § 1.84(p)(3). Corrected drawing sheets in compliance with 37 CFR 1.121(d) are required in reply to the Office action to avoid abandonment of the application. Any amended replacement drawing sheet should include all of the figures appearing on the immediate prior version of the sheet, even if only one figure is being amended. The figure or figure number of an amended drawing should not be labeled as “amended.” If a drawing figure is to be canceled, the appropriate figure must be removed from the replacement sheet, and where necessary, the remaining figures must be renumbered and appropriate changes made to the brief description of the several views of the drawings for consistency. Additional replacement sheets may be necessary to show the renumbering of the remaining figures. Each drawing sheet submitted after the filing date of an application must be labeled in the top margin as either “Replacement Sheet” or “New Sheet” pursuant to 37 CFR 1.121(d). If the changes are not accepted by the examiner, the applicant will be notified and informed of any required corrective action in the next Office action. The objection to the drawings will not be held in abeyance.
Specification
The lengthy specification has not been checked to the extent necessary to determine the presence of all possible minor errors. Applicant’s cooperation is requested in correcting any errors of which applicant may become aware in the specification.
Claim Objections
Claim 7 is objected to because of the following informalities: the semicolon at the end of the claim should be a period.
Claim 8 is objected to because of the following informalities: the comma at the end of the first limitation should be a semicolon for consistency with the remainder of the claim.
Appropriate correction is required.
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 1-10 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 "the random recurrent weight matrix" three lines from the bottom of the first limitation. There is insufficient antecedent basis for this limitation in the claim. For purposes of examination, this limitation will be construed as “the recurrent weight matrix”.
Claims 7-8 recite the limitation "said non-linear activation layer". Claim 8 further recites “said one-dimensional convolutional layer”. There is insufficient antecedent basis for these limitations in the claims. Examiner recommends that both of these claims be amended to depend on claim 3, which provides antecedent basis for both terms.
All claims dependent on a claim rejected hereunder are also rejected for being dependent on a rejected base claim.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-10 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The claim(s) does/do not fall within at least one of the four categories of patent eligible subject matter because, under their broadest reasonable interpretation in light of the specification, they are directed to software per se. Claim 1 is directed to an apparatus comprising a reservoir comprising a recurrent neural network (i.e., software) and a readout comprising a one-dimensional, temporal convolutional neural network (i.e., also software). No hardware such as a processor or a memory is recited anywhere in the claims, and there is no reason to believe that the neural networks are limited to hardware instantiations. Examiner recommends that Applicant amend the claims explicitly to recite the hardware described at, for instance, paragraphs 51-52 of the specification as originally filed.
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.
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 1-2 and 10 are rejected under 35 U.S.C. 103 as being unpatentable over Virbila et al. (US 20220222512) (“Virbila”) in view of Peng et al. (US 20220397874) (“Peng”) and further in view of Griffith et al. (WO 2021067358) (“Griffith”).
Regarding claim 1, Virbila discloses “[a]n apparatus comprising:
a reservoir comprising a recurrent neural network and receiving at least one input temporal sequence, the at least one input temporal sequence comprising a data space dimension (Virbila Fig. 1 shows a reservoir 102 that receives a temporal input; paragraph 9 discloses that the reservoir maps an input signal vector to a high-dimensional state space [i.e., the input comprises data space vector dimensions]; see also paragraph 31 (describing the reservoir as a type of RNN)), said recurrent neural network comprising an initially unlearned input weight matrix and an initially unlearned recurrent weight matrix (As is an m x m matrix specifying the set of mixing weights that govern the reservoir dynamics [i.e., it is a recurrent weight matrix], and Bs is a weight matrix that maps the input into the reservoir [i.e., it is an input weight matrix] – Virbila, paragraph 58; see also paragraph 31 (indicating that the recurrent connections are fixed, i.e., unlearned)), said recurrent neural network comprising a plurality of neurons corresponding to a plurality of reservoir activities (reservoir computer is a special form of recurrent neural network (i.e., a neural network having feedback connections between nodes [neurons each performing a reservoir-related activity]), where the recurrent connections are fixed and not adapted by the input signal – Virbila, paragraph 31), the initially unlearned input weight matrix projecting the at least one input temporal sequence from the data space dimension into a dimensionally higher reservoir space dimension (cognitive signal processor includes a reservoir computer, which accepts mixture signals as input and maps [projects] them to a high-dimensional dynamical system [i.e., higher-dimensional than the input] known as the reservoir – Virbila, paragraph 31; Bs is a weight matrix that maps the input into the reservoir [i.e., it is an input weight matrix] – id. at paragraph 58), …, said plurality of neurons receiving the projected input temporal sequence and the random recurrent weight matrix (As is an m x m matrix specifying the set of mixing weights that govern the reservoir dynamics [i.e., it is a recurrent weight matrix, and since they govern the reservoir dynamics, they are received by the reservoir] – Virbila, paragraph 58; cognitive signal processor includes a reservoir computer, which accepts mixture signals as input and maps [projects] them to a high-dimensional dynamical system known as the reservoir [i.e., the reservoir receives the projected sequence] – id. at paragraph 31), said plurality of neurons collectively outputting a plurality of reservoir state vectors (dynamic reservoir state vector is the sum of the previous reservoir state multiplied by a transition matrix and an input signal multiplied by an input-to-reservoir mapping matrix [i.e., the state is the output of these state-space equations calculated by the reservoir’s neurons] – Virbila, paragraph 57), the plurality of reservoir state vectors being stacked to form a reservoir state matrix (weight adaptation component provides weights that include the output layer, which are combined with the reservoir state matrix to obtain the final output [note that a matrix is a series of stacked vectors] – Virbila, paragraph 64); and
a readout comprising a … neural network (a reservoir has trainable readout layers that can be trained to learn desired outputs by utilizing the state functions [such layers comprising a neural network] – Virbila, paragraph 32), said … neural network receiving the reservoir state matrix from said reservoir (weight adaptation component provides as an output a matrix of weights to an output layer computer; the weight adaptation component provides weights that include the output layer, which are combined with the reservoir state matrix to obtain the final output [i.e., the output layer of the network receives the reservoir state matrix for combination with the weights to obtain the output] – Virbila, paragraph 64), … thereby respectively filtering a plurality of temporal features (reservoir computer is implemented as an adaptable state-space filter [i.e., the outputs are filtered features] – Virbila, paragraph 33; see also Fig. 1 (showing that the input is a time sequence and the outputs are features)).”
Virbila appears not to disclose explicitly the further limitations of the claim. However, Peng discloses “a one-dimensional, temporal convolutional neural network, … said one-dimensional, temporal convolutional network comprising a stack of one-dimensional convolutional blocks, said stack of one-dimensional convolutional blocks convolving the … matrix over time (extraction layer accepts a t-th input matrix and performs a one-dimensional convolution in the time dimension [i.e., convolves the input matrix] – Peng, paragraph 30; see also paragraph 40 (disclosing that the blocks are stacked)) ….”
Peng and the instant application both relate to convolutional neural networks and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Virbila to employ a one-dimensional temporal convolutional neural network, as disclosed by Peng, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would reduce the processing power used to perform the convolution by reducing the dimensionality. See Peng, paragraph 30.
Neither Virbila nor Peng appears to disclose explicitly the further limitations of the claim. However, Griffith discloses that “a number of neurons in the plurality of neurons [is] equal to a number of dimensions of the reservoir space dimension (the dynamics of the reservoir are described by an equation that is a function of a vector r(t), where each dimension of the vector r represents a single node in the network [i.e., the vector dimensionality is equal to the number of reservoir nodes] – Griffith, paragraph 37) ….”
Griffith and the instant application both relate to reservoir computing and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Virbila and Peng to employ a reservoir space dimension equal to the number of neurons, as disclosed by Griffith, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the dimensionality of the space is appropriate to the size of the reservoir. See Griffith, paragraph 37.
Regarding claim 2, Virbila, as modified by Peng and Griffith, discloses that “said recurrent neural network comprises a random, recurrent neural network (reservoir computers may contain weights in both a reservoir [RNN] connectivity matrix and an input-to-reservoir mapping vector that are chosen randomly – Virbila, paragraph 35), …
said at least one input temporal sequence comprises a plurality of input temporal sequences (cognitive signal processor includes a reservoir computer, which accepts mixture signals as input and maps the signals to the reservoir [note that the use of the plural “signals” implies that there are in general multiple signals] – Virbila, paragraph 31; see also Fig. 1 (showing that the input is a temporal sequence)), [and] …
the reservoir state matrix is in a reservoir data space, the reservoir data space comprising the reservoir space dimension (reservoir computing operates by projecting an input signal vector into a high-dimensional reservoir state space [data space comprising reservoir space dimensions] – Virbila, paragraph 32; reservoir state matrix is combined with weights to obtain a final output [i.e., the reservoir state matrix is part of the reservoir state space] – id. at paragraph 64).”
Regarding claim 10, Virbila, as modified by Peng and Griffith, discloses “a gateway directly connecting said reservoir to said readout (Virbila Fig. 1 shows that the reservoir 102 is directly connected to the trainable readouts 104).”
Claims 3-4 and 7-8 are rejected under 35 U.S.C. 103 as being unpatentable over Virbila in view of Peng and Griffith and further in view of Kollada et al. (US 20220392637) (“Kollada”).
Regarding claim 3, neither Virbila, Peng, nor Griffith appears to disclose explicitly the further limitations of the claim. However, Kollada discloses that “each one-dimensional convolutional block of said stack of one-dimensional convolutional blocks comprises a one-dimensional convolutional layer and a non-linear activation layer (multimodal classification framework comprises unimodal encoders of stacked convolution blocks (1D convolution layer followed by batch normalization and ReLU activation [non-linear activation]) – Kollada, paragraph 137).”
Kollada and the instant application both relate to machine learning and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Virbila, Peng, and Griffith to employ 1D convolutional blocks comprising a 1D convolutional layer and an activation function layer, as disclosed by Kollada, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow for the construction of more complex networks with relatively simple building blocks. See Kollada, paragraph 137.
Regarding claim 4, Virbila, as modified by Peng, Griffith, and Kollada, discloses that “said one-dimensional, temporal convolutional network comprises:
a fully connected layer connected to said stack of one-dimensional convolutional blocks (output of the 1D convolutional block is flattened through max and mean pooling, and the two vectors are concatenated and input to a fully-connected layer followed by sigmoid activation – Kollada, paragraph 93).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Virbila, Peng, and Griffith to employ a fully-connected layer in the architecture, as disclosed by Kollada, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow for the construction of more complex networks with relatively simple building blocks. See Kollada, paragraph 137.
Regarding claim 7, Virbila, as modified by Peng, Griffith, and Kollada, discloses that “said non-linear activation layer comprises one of:
a Rectified Linear Unit function;
a leaky Rectified Linear Unit function;
a Gaussian Error Linear Unit function;
a Sigmoid function;
a Softmax function; and
a tanh function (multimodal classification framework comprises unimodal encoders of stacked convolution blocks (1D convolution layer followed by batch normalization and ReLU [i.e., rectified linear unit] activation) – Kollada, paragraph 137)[.]” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Virbila, Peng, and Griffith to employ 1D convolutional blocks comprising a 1D convolutional layer and a ReLU layer, as disclosed by Kollada, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow for the construction of more complex networks with relatively simple building blocks. See Kollada, paragraph 137.
Regarding claim 8, Virbila, as modified by Peng, Griffith, and Kollada, discloses that “said each one-dimensional convolutional block comprises one of:
a pooling layer between said one-dimensional convolutional layer and said non-linear activation layer[;]
a downsampling layer; and
a batch normalization layer between said one-dimensional convolutional layer and said non-linear activation layer (multimodal classification framework comprises unimodal encoders of stacked convolution blocks (1D convolution layer followed by batch normalization and ReLU [i.e., rectified linear unit] activation) – Kollada, paragraph 137).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Virbila, Peng, and Griffith to employ 1D convolutional blocks comprising a 1D convolutional layer, a batch normalization layer, and a ReLU layer, as disclosed by Kollada, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow for the construction of more complex networks with relatively simple building blocks. See Kollada, paragraph 137.
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Virbila in view of Peng and Griffith and further in view of Kollada and Keum et al. (US 20230385656) (“Keum”).
Regarding claim 5, neither Virbila, Peng, Griffith, nor Kollada appears to disclose explicitly the further limitations of the claim. However, Keum discloses that “said fully connected layer comprises one of a many-to-one classifier, a one-to-many classifier, and a many-to-many classifier (in the case of a one-to-many model, some layers are shared between classes, and the fully connected layer is provided for each class – Keum, paragraph 44).”
Keum and the instant application both relate to neural networks and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Virbila, Peng, Griffith, and Kollada to employ a one-to-many classifier, as disclosed by Keum, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would simplify the inference processing by making the classification decision merely whether the input falls into a particular class or not. See Keum, paragraph 44.
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Virbila in view of Peng and Griffith and further in view of Kollada and Lee et al. (US 20220114431) (“Lee”).
Regarding claim 6, neither Virbila, Peng, Griffith, nor Kollada appears to disclose explicitly the further limitations of the claim. However, Lee discloses that “said fully connected layer comprises a perceptron (sequence of neural-net component layers includes both convolutional layers and fully-connected (perceptron) layers in configurable quantities – Lee, paragraph 22).”
Lee and the instant application both relate to neural networks and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Virbila, Peng, Griffith, and Kollada to employ a perceptron as the fully-connected layer, as disclosed by Lee, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow the user to use a well-known, of-the-shelf model without having to develop novel architectures. See Lee, paragraph 22.
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over Virbila in view of Peng and Griffith and further in view of Kollada and Tung et al. (US 11636337) (“Tung”).
Regarding claim 9, neither Virbila, Peng, Griffith, nor Kollada appears to disclose explicitly the further limitations of the claim. However, Tung discloses that “said downsampling layer comprises a strided convolution layer (downsampling is performed by strided convolutions in the first layers of conv3 and conv4 – Tung, col. 9, ll. 23-41).”
Tung and the instant application both relate to neural networks and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Virbila, Peng, Griffith, and Kollada to perform downsampling by strided convolution, as disclosed by Tung, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would force the network to focus on the most discriminative features, thereby ignoring information that is redundant. See Tung, col. 9, ll. 23-41.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to RYAN C VAUGHN whose telephone number is (571)272-4849. The examiner can normally be reached M-R 7:00a-5:00p ET.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Kamran Afshar, can be reached at 571-272-7796. 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.
/RYAN C VAUGHN/ Primary Examiner, Art Unit 2125
Prosecution Insights