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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 02/17/2026 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Status of Claims
The present application is being examined under the claims filed 02/17/2026. The status of the claims are as follows:
Claims 1-24 are pending;
Claims 1-5, 7-13, 15-21, 23, and 24 are amended;
Claims 6, 14, and 22 are canceled.
Response to Amendments
The Office action is in response to Applicant’s communication filed 02/17/2026 in response to the Office action mailed 10/14/2025. Applicant’s remarks and amendments to the claims have been considered with the results set forth below.
Upon reconsideration, the prior rejection under 35 U.S.C. § 101 is withdrawn. The prior rejection under 35 U.S.C. § 112(a) is withdrawn. Claims 1, 7-9, 15-17, and 23-24 remain rejected under 35 U.S.C. § 112(b) for the reasons set forth below. Claims 1-5, 7-13, 15-21, 23, and 24 remain rejected under 35 U.S.C. § 103 for the reasons set forth below.
Response to Arguments
Regarding 35 U.S.C. § 101
The Examiner has considered Applicant’s arguments and amendments regarding the rejection of claims 1-20 under 35 U.S.C. § 101 and finds them persuasive.
Applicant argues that the amended claims are directed to a specific multistage classification architecture that reduces computational burden by processing lower-dimensional vector series using second system models before completing processing with the first system model. Upon reconsideration, the Examiner agrees that the amended claims now recite a more particular ordered combination directed to staged lower-dimensional classification processing rather than merely reciting classification at a high level of generality.
Accordingly, the rejection of claims 1-24 under 35 U.S.C. § 101 is withdrawn.
Regarding 35 U.S.C. § 112
The Examiner has considered Applicant’s arguments and amendments regarding the rejection of claims 1-24 under 35 U.S.C. § 112 and finds them persuasive in part.
Applicant argues that the prior §112 issues concerning “stage knowledge map” and “identification stage” are overcome by the amendments and by the cited support in the specification. These arguments are persuasive with respect to the written-description rejection. The record supports that the specification describes knowledge-map terminology and identification-stage concepts sufficiently for purposes of the prior §112(a) rejection.
Accordingly, the prior rejection of claims 2, 5, 10, 13, 18, and 21 under 35 U.S.C. § 112(a) is withdrawn.
However, Applicant’s amendments do not place the claims in condition for allowance because the claims remain indefinite under 35 U.S.C. § 112(b) for the reasons set forth below. In particular, the amended independent claims recite the phrase “no one or more”, which renders the conditional logic unclear. Claims 7, 15, and 23 recite “related stages” and “neighboring classes” without objective boundaries. Claims 8, 16, and 24 recite a “probability of substantially one”, which is a term of degree lacking an objective standard for determining the requisite/necessary degree.
The §112(b) rejection below are therefore maintained/newly set forth as applicable.
Regarding 35 U.S.C. § 103
The Examiner has considered Applicant’s arguments and amendments regarding the rejection of claims 1-24 under 35 U.S.C. § 103 and finds them unpersuasive.
Applicant argues that Jones and Baker do not teach or suggest determining a plurality of second series of vectors having fewer dimensions than the first series of input vectors by omitting one or more dimensions, constructing a plurality of training vectors series based on the plurality of second series of vectors, constructing a plurality of second system models, and using the second system models as a screening stage before completing processing with the first system model.
These arguments are not persuasive.
As an initial matter, the amended claims do not require a particular dimensionality-reduction algorithm, do not require that the omitted dimensions be permanently discarded, and do not require that the second system models be generated using any specific mathematical feature-selection process. The claims broadly recite determining lower-dimensional second series of vectors by omitting one or more dimensions of the first series of input vectors and using second system models to determine whether the second series of vectors is identifiable as belonging to a class before completing processing with the first system model.
Baker teaches a multistage machine-learning recognition system in which non-final-stage machine-learning classifiers assign each data input to one or more, but less than all, final-stage machine-learning classifiers. Baker explains that this architecture improves computational efficiency because only a small fraction of late-stage machine-learning systems receive each item of data. See Baker, ¶¶ [0003], [0024]-[0025], and claims 1 and 3.
Jones teaches a cascaded classifier architecture in which early classifiers evaluate relatively few features and reject candidates before later, more accurate classifiers perform further processing. Jones further teaches that AdaBoost selects a small set of features for each classifier and that early classifiers may evaluate only one or two features while later classifiers evaluate more features. See Jones, ¶¶ [0067]-[0082].
Thus, Baker teaches the claimed staged machine-learning architecture and the computational-efficiency rationale, while Jones teaches using selected subsets of features in early classifiers to reduce processing and reject candidates before further processing. It would have been obvious to implement Baker’s non-final-stage classifiers using Jone’s reduced-feature early classifier technique so that the non-final stage processes lower-dimensional feature representations and avoids invoking the first/final system model unless preliminary identifiability is indicated.
Applicant’s argument that Jones does not expressly use the phrase “omitting dimensions” is not persuasive because Jones teaches selecting and evaluating only a subset of available features for a classifier. Treating available features as dimensions of an input representation, the non-selected features correspond to omitted dimensions for that classifier’s effective input representation. Jones therefore teaches or at least suggests forming reduced-dimensional classifier inputs by omitting non-selected feature dimensions.
Applicant’s argument that Baker does not teach lower-dimensional vector construction is also not persuasive against the combination. Baker is relied upon primarily for the multistage machine-learning architecture in which non-final stages route data to selected later-stage classifiers for computational efficiency. Jones is relied upon for the reduced-feature classifier technique in Baker’s known multistage classifier-routing architecture to predictably reduce computational burden before invoking more complete downstream model processing.
Accordingly, the rejections of claims 1-5, 7-13, 15-21, 23, and 24 under 35 U.S.C. § 103 is maintained in modified form as set forth below.
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 1-5, 7-13, 15-21, and 23-24 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.
Claims 1, 9, and 17 recite, in relevant part, “in response to determining that no one or more of the plurality of second system models indicates that one or more of the plurality of second series of vectors is identifiable as belonging to a class.”
The phrase “no one or more” renders the scope of the claims unclear. Under the broadest reasonable interpretation in light of the specification, it is unclear whether this condition is met when none of the plurality of second system models indicates identifiability, when not more than one second system model indicates identifiability, or under some other condition. This ambiguity is material because the recited condition controls the branch in which the system determines that the first series of input vectors does not belong to a class and causes a corresponding action to be taken. Accordingly, one of ordinary skill in the art would not be reasonably apprised of the scope of the conditional logic recited in the claims.
Dependent claims 2-5, 7, 8, 10-13, 15, 16, 18-21, 23, and 24 depend from and include the limitations of independent claims 1, 9, or 17. Accordingly, these dependent claims contain the same deficiencies identified above with respect to independent claims 1, 9, and 17 and are rejected under 35 U.S.C. § 112(b) for the same reasons.
Claims 7, 15, and 23 recite determining whether “related stages” have been exercised and continuing to iterate if not all “related stages” have been exercised.
The phrase “related stages” renders the scope of the claims unclear. Under the broadest reasonable interpretation in light of the specification, the claims do not identify what relationship makes one stage “related” to another stage for purposes of determining when iteration is complete. For example, the claims do not specify whether stages are related based on a common hierarchy branch, common candidate class, common subset of vector dimensions, common system model, adjacency in a stage sequence, or another objective criterion. Because the “related stages” determination controls whether the system continues iterating or proceeds to probability determination, one of ordinary skill in the art would not be reasonably apprised of the scope of the claimed stage relationship.
Claims 7, 15, and 23 recite determining a probability of the input vector belonging to one or more classes “based on neighboring classes”.
The phrase “neighboring classes” renders the scope of the claims unclear. Under the broadest reasonable interpretation in light of the specification, the claims do not identify how classes are determined to be “neighboring”. Classes, as recited, are categories or labels and do not inherently have spatial, topological, hierarchical, or metric proximity absent a recited relationship. It is unclear whether “neighboring classes” refers to classes associated with nearest knowledge elements, class centroids in vector space, adjacent nodes in a hierarchy, overlapping influence fields, or another objective relationship. Accordingly, one of ordinary skill in the art would not be reasonably apprised of the scope of “neighboring classes”.
The term “substantially one” in claims 8, 16, and is a relative term which renders the claim indefinite. The term “substantially one” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention.
In particular, claims 8, 16, and 24 recite determining that the input vector belongs to the class with a “probability of substantially one”. The claims do not specify a numerical threshold, tolerance, or objective condition under which a probability is “substantially one” rather than merely high. It is therefore unclear whether the claimed probability must be equal to 1, at least 0.99, at least 0.95, at least 0.90, or satisfy some other certainty condition.
Claim Objections
Claims 1, 9, and are objected to because of the following informalities:
"number of dimension" should be amended to "number of dimensions", if intended.
“training vectors series” should be amended to “training vector series”, “series of training vectors”, or another grammatically correct formulation, if intended.
“does not belong to a class and cause a corresponding action to be taken” should be amended for grammatical consistency, if intended.
Appropriate correction is required.
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.
Claims 1, 3, 4, 8, 9, 11, 12, 16, 17, 19, 20, and 24 are rejected under 35 U.S.C. 103 as being unpatentable over James K. Baker (US20200051550A1) in view of Yuchun Tang (US8160975B2) and further in view of Michael J. Jones (US20020102024A1).
Regarding claim 1, Baker, in view of Tang and further in view of Jones, teaches a multistage classification system comprising: a memory; one or more processors; and logic operable to cause the one or more processors to:
“obtain a first series of input vectors representing ” — Baker teaches this limitation in part. Baker discloses that machine learning is implemented by computers to make predictions on data by:
“building models from sample data inputs.” (Baker, pg. 1, ¶[0002])
Baker further discloses:
“classifying a data input to a classification output;” (Baker, pg. 13, ¶[0107]; pg. 14, claim 1)
and that data inputs are assigned to final-stage classifiers:
“data assignment classifiers ... may assign each data input to the one or more of, and less than N of, the N final-stage machine learning classifiers” (Baker, pg. 13, ¶[0107])
Baker therefore teaches obtaining input data for machine-learning classification and processing the data using a multistage classifier system. Baker does not expressly teach that the input vectors represent “a series of measurements of one or more physical objects.” Jones teaches that portion, as set forth below.
“establish a first system model based on the first series of input vectors” — Baker teaches this limitation. Baker discloses:
“N final-stage machine learning classifiers, wherein N>1,” (Baker, pg. 13, ¶[0107])
where:
“each of the N final-stage machine learning classifiers is for classifying a data input to a classification output.” (Baker, pg. 13, ¶[0107])
Baker further discloses:
“building models from sample data inputs.” (Baker, pg. 1, ¶[0002])
Baker’s final-stage machine-learning classifier corresponds to the claimed first system model because it is a machine-learning model built from sample data inputs and used to classify a data input to a classification output.
“determine a plurality of second series of vectors based on the first series of input vectors” – Baker teaches this limitation in part. Baker discloses:
“at least one non-final stage” (Baker, pg. 12, ¶[0107])
where each non-final stage includes:
“one or more machine learning, data assignment classifiers that assigns each data input to one or more of, and less than N of, the N final-stage machine learning classifiers.” (Baker, pg. 13, ¶[0107])
Baker further discloses embodiments having:
“M second stage machine learning classifiers, where M>1” (Baker, pg. 13, ¶[0109])
where each second-stage classifier classifies each data input to one or more but less than N final-stage classifiers. Baker therefore teaches plural intermediate/second classifiers that process data prior to final-stage classification.
“process the plurality of second series of vectors using a learning system model, including:” – Baker teaches this limitation at least in part. Baker discloses:
“at least one non-final stage” (Baker, pg. 13, ¶[0107])
where each non-final stage includes:
“one or more machine learning, data assignment classifiers that assigns each data input to one or more of, and less than N of, the N final-stage machine learning classifiers.” (Baker, pg. 13, ¶[0107])
Baker further discloses:
“M second stage machine learning classifiers, where M>1” (Baker, pg. 13, ¶[0109])
Thus, Baker teaches processing data using non-final and second-stage machine-learning classifiers before final-stage classification.
“construct, ” — Baker teaches this limitation at least in part by teaching plural non-final/second-stage machine-learning classifiers. Baker discloses:
“one or more machine learning, data assignment classifiers” (Baker, pg. 13, ¶[0108])
and further discloses:
“M second stage machine learning classifiers, where M>1.” (Baker, pg. 13, ¶[0109])
Baker therefore teaches plural second/intermediate system models. Baker does not expressly teach constructing those second system models using reduced-dimensional training vector series. Tang teaches that portion, as set forth below.
“process, using the plurality of second system models, the plurality of second series of vectors” — Baker teaches this limitation at least in part. Baker discloses:
“one or more machine learning, data assignment classifiers that assigns each data input to one or more of, and less than N of, the N final-stage machine learning classifiers.” (Baker, pg. 13, ¶[0107])
Baker further discloses:
“M second stage machine learning classifiers, where M>1.” (Baker, pg. 13, ¶[0109])
Baker therefore teaches processing data using plural non-final or second-stage machine-learning classifiers.
“determine whether one or more of the plurality of second system models indicates that one or more of the plurality of second series of vectors is identifiable as belonging to a class” — Baker teaches this limitation at least in part. Baker discloses that a machine-learning data-assignment classifier assigns each data input to one or more final-stage classifiers:
“that the machine learning, data assignment classifier determines will classify the data input correctly.” (Baker, pg. 13, ¶[0108])
Thus, Baker teaches determining whether one or more non-final or second-stage classifiers indicates that a data input should be processed by selected downstream classifiers that are expected to classify the input correctly.
“in response to determining that one or more of the plurality of second system models indicates that one or more of the plurality of second series of vectors is identifiable as belonging to a class, complete processing of the first series of input vectors in association with the first system model” — Baker teaches this limitation. Baker discloses that non-final-stage classifiers assign each data input to:
“one or more of, and less than N of, the N final-stage machine learning classifiers.” (Baker, pg. 13, ¶[0107])
Baker further discloses:
“each of the N final-stage machine learning classifiers is for classifying a data input to a classification output.” (Baker, pg. 13, ¶[0107])
Thus, when Baker’s non-final-stage classifier determines that selected final-stage classifiers should receive the data input, processing is completed by the selected final-stage classifier, corresponding to the claimed first system model.
“determine whether the first system model indicates that the first series of input vectors belongs to a class and ” — Baker teaches this limitation at least in part. Baker discloses that each final-stage machine-learning classifier is:
“for classifying a data input to a classification output.” (Baker, pg. 13, ¶[0107])
Classification of the data input to a classification output corresponds to determining whether the first system model indicates that the first series belongs to a class. Baker does not expressly teach the portion “cause a corresponding action to be taken.” Jones teaches that portion, as set forth below.
Baker does not expressly teach these limitations and/or portions of limitations:
“…a series of measurements of one or more physical objects…”
“each second series of vectors having fewer dimensions than the number of dimension of the first series of input vectors by omitting one or more dimensions of the first series of input vectors.”
“... using the plurality of training vectors series, ...”
“construct a plurality of training vectors series based on the plurality of second series of vectors, each training vectors series having fewer dimensions than the number of dimensions of the first series of input vectors.”
“…and cause a corresponding action to be taken”
“in response to determining that no one or more of the plurality of second system models indicates that one or more of the plurality of second series of vectors is identifiable as belonging to a class, determine that the first series of input vectors does not belong to a class and cause a corresponding action to be taken.”
Tang, however, teaches these limitations and/or portions of limitations:
“each second series of vectors having fewer dimensions than the number of dimension of the first series of input vectors by omitting one or more dimensions of the first series of input vectors” — Tang teaches this limitation. Tang discloses that support vector machine data may be plotted into:
“an n-dimensional space, n being the number of attributes associated with the item to be classified,” (Tang, col. 1, lines 11-13)
and recognizes that:
“large numbers of attributes and a large volume of training data” (Tang, col. 1, lines 14-15)
make support vector machines:
“processor intensive.” (Tang, col. 1, line 15)
Tang further teaches that:
“the in-bag tuples can be visualized as datapoints plotted onto an n-dimensional space, where n equals the number of attributes associated with each tuple.” (Tang, col. 5, lines 18-20)
Tang expressly states:
“If a dimension is removed, the datapoints can be said to be projected into the subspace comprising the remaining attributes.” (Tang, col. 5, 21-23)
Tang further states that the random subspace may be selected by:
“randomly selecting attributes to remove from the subspace or randomly selecting the attributes that are included in the subspace.” (Tang, col. 5, lines 24-26)
Thus, Tang teaches forming lower-dimensional second vector data by omitting one or more dimensions/attributes from an original n-dimensional representation.
“construct a plurality of training vectors series based on the plurality of second series of vectors, each training vectors series having fewer dimensions than the number of dimensions of the first series of input vectors” — Tang teaches this limitation. Tang discloses:
“receiving a training dataset comprising a plurality of tuples and a plurality of attributes for each of the tuples” (Tang, col. 1, lines 45-47)
and:
“deriving a plurality of granules from the training dataset, each granule comprising a plurality of sample tuples and a plurality of sample attributes.” (Tang, col. 1, lines 47-50)
Tang further discloses that:
“random selection of tuples with a projection of the tuple attributes into a random subspace generates a granule” (Tang, col. 3, lines 31-33)
and that the granule is:
“a matrix having a number of rows of records” and “a number of columns defining attributes associated with the granule.” (Tang, col. 3, lines 34-37)
Tang also states that:
“an original matrix associated with the training dataset can be reduced into a granule.” (Tang, col. 5, lines 30-32)
Therefore, Tang teaches constructing plural reduced-dimensional training vector series/granules from selected attributes of an original training dataset.
“... using the plurality of training vectors series, ...” — Tang teaches this limitation. Tang discloses:
“processing the granules using a support vector machine process to identify a hyperplane classifier associated with each of the granules.” (Tang, col. 1, lines 50-52)
Tang further discloses that:
“The processing module 220 can be operable to process the granules using a support vector machine process.” (Tang, col. 5, lines 37-38)
and that the processing module can:
“identify a hyperplane classifier.” (Tang, col. 5, line 43)
Each Tang hyperplane classifier associated with a respective reduced-dimensional granule corresponds to a claimed second system model because it is a classifier/model constructed using a corresponding reduced-dimensional training vector series.
“process, using the plurality of second system models, the plurality of second series of vectors” — Tang teaches this limitation. Tang discloses that:
“The hyperplane classifiers … can then be used to analyze new data.” (Tang, col. 3, lines 56-57)
Tang further discloses that:
“The prediction module 230 can compare datapoints associated with the features against each of the hyperplane classifiers derived from the granules to derive granule predictions associated with the respective hyperplane classifiers.” (Tang, col. 6, lines 9-12)
Tang also discloses that the prediction module may:
“plot the unclassified new tuple onto a random subspace associated with a first granule and associated hyperplane classifier.” (Tang, col. 6, lines 14-16)
Thus, Tang teaches processing reduced-dimensional input data using plural reduced-dimensional classifiers.
“determine whether one or more of the plurality of second system models indicates that one or more of the plurality of second series of vectors is identifiable as belonging to a class” — Tang teaches this limitation. Tang discloses that the hyperplane classifiers may be used on an unknown datapoint:
“to provide a plurality of predictions” (Tang, Abstract)
and that those predictions may be aggregated:
“to provide a final prediction associated with the datapoint.” (Tang, Abstract)
Tang further discloses that the prediction module may determine whether an unclassified new tuple shows characteristics associated with:
“a first classification” (Tang, col. 6, line 18)
or:
“a second classification.” (Tang, col. 6, line 19)
Thus, Tang teaches determining whether one or more reduced-dimensional classifiers indicates that an input is identifiable as belonging to a class.
Neither Tang nor Baker teaches these remaining limitations and/or portions of limitations:
“…a series of measurements of one or more physical objects…”
“…and cause a corresponding action to be taken”
“in response to determining that no one or more of the plurality of second system models indicates that one or more of the plurality of second series of vectors is identifiable as belonging to a class, determine that the first series of input vectors does not belong to a class and cause a corresponding action to be taken.”
Jones, however, teaches these remaining limitations and/or portions of limitations:
“…a series of measurements of one or more physical objects…” — Jones teaches this portion of the limitation. Jones discloses an object detector using a cascade of classifiers to classify image subwindows as to whether each subwindow is likely to contain an object. Jones discloses:
“The object detector uses a cascade of homogenous classification functions or classifiers to classify the subwindows as to whether each subwindow is likely to contain an instance of the object.” (Jones, Abstract)
Jones further discloses output processing involving detected object instances:
“output processing such as highlighting the instances 38 of objects in an output image 34 for viewing by an end-user.” (Jones, pg. 6, ¶[0067])
Thus, Jones teaches input data corresponding to measurements of physical objects represented in image data.
“…and cause a corresponding action to be taken” — Jones teaches this portion of the limitation. Jones discloses that after retained subwindows are passed through the cascade, further processing may include:
“output processing such as highlighting the instances 38 of objects in an output image 34 for viewing by an end-user.” (Jones, pg. 6, ¶[0067])
Highlighting detected object instances in an output image corresponds to causing a corresponding action to be taken.
“in response to determining that no one or more of the plurality of second system models indicates that one or more of the plurality of second series of vectors is identifiable as belonging to a class, determine that the first series of input vectors does not belong to a class and cause a corresponding action to be taken” — Jones teaches this limitation. This limitation is addressed to the extent it can be understood in view of the §112(b) rejection above. Jones discloses a cascade of classifiers used to classify subwindows as to whether each subwindow is likely to contain an object:
“The object detector uses a cascade of homogenous classification functions or classifiers to classify the subwindows as to whether each subwindow is likely to contain an instance of the object.” (Jones, Abstract)
Jones further discloses that the initial classifier:
“eliminates a large number of negative examples” (Jones, pg. 6, ¶[0067])
with:
“very little processing” (Jones, pg. 6, ¶[0067])
Jones also discloses that rejected subwindows are indicated as “False” and retained windows as “True”:
“this elimination process 88 is indicated by the letter ‘F’ for ‘False’, and the rejected set of subwindows 84 indicates those subwindows 42 eliminated by the classifiers 30-1, 30-2, 30-3, and 30-4, respectively. The retention 90 of windows is indicated by the letter ‘T’ for ‘True’, as the classifiers 30-1, 30-2, 30-3, and 30-4 pass on subwindows 42 that are not eliminated.” (Jones, pg. 6, ¶[0067])
Jones further discloses that:
“the retained subwindows 42 are passed on for further processing” (Jones, pg. 6, ¶[0067])
while rejected subwindows are eliminated:
“eliminate additional negatives (e.g., rejected subwindows 84)” (Jones, pg. 6, ¶[0068])
Thus, Jones teaches determining that a preliminary classifier does not indicate identifiability and taking a corresponding action by rejecting/eliminating the candidate rather than forwarding it for further processing.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Baker’s multistage machine-learning recognition system to use Tang’s reduced-dimensional random-subspace granule classifiers as the non-final or intermediate classifiers. Baker teaches that non-final stages direct data toward selected final-stage machine-learning systems, but less than all final-stage systems, and Baker’s claim 1 expressly recites assigning each data input to one or more but less than N final-stage classifiers. Tang teaches that large numbers of attributes and large training datasets make SVMs processor intensive and teaches reducing an original matrix into granules by projecting sample tuples into random subspaces with selected/removed attributes. A person of ordinary skill in the art would have been motivated to use Tang’s reduced-dimensional granule classifiers in Baker’s non-final stage to reduce the amount of data/classifier processing performed before selected final-stage classification.
It would further have been obvious to use Jones’s rejection/retention cascade logic in the Baker/Tang system. Jones teaches that early classifiers using a limited number of features eliminate many negative subwindows before later classifiers requiring additional computation, and that retained windows are passed on for further processing. Jones further teaches that the cascade is:
“significantly faster” (Jones, pg. 6, ¶[0071])
and has:
“much better classification performance.” (Jones, pg. 6, ¶[0071])
A person of ordinary skill would have been motivated to apply this known early-rejection technique to Baker/Tang’s multistage reduced-dimensional classifiers to avoid unnecessary final-stage processing when the preliminary reduced-dimensional models do not indicate identifiability.
Regarding claim 3, Baker in view of Tang and further in view of Jones teaches the multistage classification system of claim 1,
“wherein the processing using the learning system model is configured to continue iteratively until a classification of input vectors is achieved.” – Baker teaches this limitation. Baker discloses a multistage system in which data is processed through non-final and final-stage machine-learning classifiers. Baker discloses:
“at least one non-final stage” (Baker, pg. 13, ¶[0107])
where each non-final stage includes:
“one or more machine learning, data assignment classifiers that assigns each data input to one or more of, and less than N of, the N final-stage machine learning classifiers.” (Baker, pg. 13, ¶[0107])
Baker further discloses:
“M second stage machine learning classifiers, where M>1” (Baker, pg. 13, ¶[0109])
and that each second-stage classifier:
“classifies each data input to one or more of, and less than N of, the N final stage machine learning classifiers.” (Baker, pg. 13, ¶[0109])
Baker further discloses that each final-stage classifier is:
“for classifying a data input to a classification output.” (Baker, pg. 13, ¶[0107])
Thus, Baker teaches processing data through one or more non-final/second-stage classifiers and then through selected final-stage classifiers until a classification output is achieved. Therefore, Baker teaches that the processing using the learning system model is configured to continue iteratively until a classification of input vectors is achieved.
Regarding claim 4, Baker in view of Tang and further in view of Jones teaches the multistage classification system of claim 3, wherein:
“iterations of the learning system model – Baker teaches this limitation in part. Baker discloses embodiments having:
“M second stage machine learning classifiers, where M>1” (Baker, pg. 13, ¶[0109])
where:
“each of the M second stage machine learning classifiers of the second non-final stage classifies each data input to it to one or more of, and less than N of, the N final stage machine learning classifiers of the final stage.” (Baker, pg. 13, ¶[0109])
Baker does not expressly teach that the iterations use:
“... use different elements of an input vector.”
Tang, however, teaches this portion of the limitation:
“... use different elements of an input vector.” – Tang teaches this limitation. Tang discloses:
“the random subspace can be selected by randomly selecting attributes to remove from the subspace or randomly selecting the attributes that are included in the subspace.”(Tang, col. 5, lines 23-26)
Tang further discloses:
“If a dimension is removed, the datapoints can be said to be projected into the subspace comprising the remaining attributes.” (Tang, col. 5, lines 20-23)
Tang also discloses:
“Granules can be continued to be selected until a threshold number of granules have been selected.” (Tang, col. 5, lines 32-33)
and:
“The random selection of the granules and a smaller sample size can facilitate diversity among the granules. For example, one granule is unlikely to be similar to any of the other granules.” (Tang, col. 5, lines 33-36)
Thus, Tang teaches that different granules/random subspaces are generated using different selected attributes, i.e., different elements/dimensions of an input vector. When applied to Baker’s iterative/staged learning system model, Tang teaches or suggests that iterations of the learning system model use different elements of an input vector.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to configure the Baker/Tang/Jones system such that iterations use different elements of an input vector, as taught by Tang. Baker teaches staged/iterative processing through plural non-final and final-stage classifiers, and Tang teaches constructing plural random-subspace granules by selecting different attributes/dimensions from the input data. A person of ordinary skill in the art would have been motivated to use different vector elements in different iterations to provide diverse reduced-dimensional classifiers and reduce processing complexity while maintaining classification performance.
Regarding claims 9, 11, 12, 16, 17, 19, 20, and 24 (non-transitory computer-readable medium claims)
Claims 9, 11, 12, and 16 are rejected under 35 U.S.C. §103 as being unpatentable over Baker in view of Tang and further in view of Jones, for the same reasons set forth for corresponding system claims 1, 3, 4, and 8, respectively.
Each of claims 9, 11, 12, and 16 recites a non-transitory computer-readable medium storing instructions to cause one or more processors to perform substantially the same functions recited in corresponding system claims 1, 3, 4, and 8. Each of claims 17, 19, 20, and 24 recites computer-implemented method steps corresponding to the functions recited in corresponding system claims 1, 3, 4, and 8.
Baker teaches the multistage machine-learning architecture including non-final-stage data-assignment classifiers and final-stage machine-learning classifiers, where data inputs are assigned to one or more but less than all final-stage classifiers, as set forth above with respect to claims 1, 3, and 8.
Tang teaches reduced-dimensional vector/training data generated by selecting or removing attributes/dimensions, constructing granules from sample tuples and sample attributes, and constructing hyperplane classifiers from those granules, as set forth above with respect to claims 1 and 4.
Jones teaches object-detection input data corresponding to physical objects, cascade-stage rejection/retention, elimination of negative subwindows, retained subwindows being passed on for further processing, and output processing such as highlighting detected object instances, as set forth above with respect to claims 1 and 8.
Expressing the same computer-implemented operations as stored instructions on a non-transitory computer-readable medium or as method steps does not, without more, impart patentable distinction over the previously established system combination. See MPEP §§2113 and 2114.
Accordingly, claims 9, 11, 12, 16, 17, 19, 20, and 24 are obvious over the cited combinations for the same reasons discussed for corresponding system claims 1, 3, 4, and 8.
Claims 2, 5, 7, 10, 13, 15, 18, 21, and 23 are rejected under 35 U.S.C. 103 as being unpatentable over Baker in view of Tang, further in view of Jones, and further in view of Tyson J. Thomas (US11461683B2).
Regarding claim 2, Baker, in view of Tang, further in view of Jones, and further in view of Thomas, teaches the multistage classification system of claim 1, the logic operable to cause the one or more processors to perform one or more of:
“identifying a subset of input vector dimensions;” – Baker does not teach this limitation. Tang, however, teaches this limitation. Tang discloses:
“the random subspace can be selected by randomly selecting attributes to remove from the subspace or randomly selecting the attributes that are included in the subspace.” (Tang, col. 5, lines 23-26)
Tang further discloses:
“If a dimension is removed, the datapoints can be said to be projected into the subspace comprising the remaining attributes.” (Tang, col. 5, lines 20-23)
Thus, Tang teaches identifying a subset of input vector dimensions because Tang randomly selects attributes to include in the subspace or attributes to remove from the subspace, thereby identifying the dimensions used for the reduced-dimensional vectors.
Neither Tang nor Baker teaches these limitations and/or portions of:
“constructing a stage knowledge map;”
“storing the identified subset of input vector dimensions and the stage knowledge map in association with an identification stage.”
Thomas, however, teaches these limitations and/or portions of:
“constructing a stage knowledge map;” – Thomas discloses:
“a pattern recognition engine maintains a knowledge element array which is a memory space for one or more knowledge maps.” (Thomas, col. 21, lines 30-34)
Thomas further discloses:
“Each knowledge map includes one or more knowledge elements, which itself includes a vector, and a category identifier.” (Thomas, col. 21, lines 32-34)
Thomas also discloses that:
“An overall pattern recognition process may be defined or configured as a series or set of individual pattern recognition operations, each associated with a configured partition.” (Thomas, col. 21, lines 59-61)
and that the partitions may be arranged:
“in a serial or hierarchical relationship.” (Thomas, col. 21, line 64)
Thus, Thomas teaches constructing knowledge maps associated with particular configured partitions/recognition operations in a staged pattern-recognition process. The knowledge map associated with a particular configured partition or recognition operation corresponds to the claimed stage knowledge map.
“storing the identified subset of input vector dimensions and the stage knowledge map in association with an identification stage.” – Thomas discloses that:
“A partition may include one or more of the following configuration parameters: 1) number of vector operands; 2) vector operand type; 3) vector operand width; 4) comparison technique; 5) distance calculation technique; and 6) maximum number of knowledge elements.” (Thomas, col. 21, lines 7-11)
Thomas further discloses:
“A pattern recognition application can invoke a particular partition by identifying the partition when passing a learn, configure, or recognize command to the knowledge element array.” (Thomas, col. 21, lines 48-51)
and that:
“An overall pattern recognition process may be defined or configured as a series or set of individual pattern recognition operations, each associated with a configured partition.” (Thomas, col. 21, lines 59-61)
Thomas further discloses:
“one recognition stage could use a first partition to provide a result.” (Thomas, col. 25, lines 49-50)
and:
“a subsequent recognition stage, using the same or a different input vector, could be performed in a different partition based on the opaque user data returned by the first recognition stage.” (Thomas, col. 25, lines 49-53)
Thus, Thomas teaches associating a knowledge map/partition configuration, including vector operand parameters, with a recognition stage in a serial or hierarchical pattern-recognition process. In view of Tang’s identified subset of vector dimensions and Thomas’s partition configuration including vector operands and associated knowledge maps, it would have been obvious to store the selected subset of input vector dimensions with the stage knowledge map for the corresponding identification stage so that the system can later invoke the proper partition/knowledge map and apply the correct vector-dimension configuration during staged recognition.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to further modify the Baker/Tang/Jones system to implement the classifiers/models using Thomas’s knowledge-map/partition structure. Baker teaches staged machine-learning classifiers, Tang teaches reduced-dimensional input/training data produced by selecting/removing attributes, and Jones teaches staged rejection/retention processing. Thomas teaches a known pattern-recognition implementation in which knowledge maps include knowledge elements having vectors and category identifiers, and in which recognition operations are organized as configured partitions arranged in serial or hierarchical stages. A person of ordinary skill in the art would have been motivated to use Thomas’s knowledge-map partition structure for the Baker/Tang/Jones staged classifier system to maintain the vector/category knowledge used by each stage and to associate the appropriate vector configuration with the corresponding stage for learning and recognition operations.
Regarding claim 5, Baker, in view of Tang, further in view of Jones, and further in view of Thomas, teaches teaches the multistage classification system of claim 1, the logic operable to cause the one or more processors to perform one or more of:
“retrieving a subset of input vector dimensions — Baker does not expressly teach this limitation. Tang and Thomas, however, teach this limitation and/or portions thereof. Tang teaches identifying/retrieving the subset of input vector dimensions. Tang discloses:
“the random subspace can be selected by randomly selecting attributes to remove from the subspace or randomly selecting the attributes that are included in the subspace.” (Tang, col. 5, lines 23-26)
Tang further discloses:
“If a dimension is removed, the datapoints can be said to be projected into the subspace comprising the remaining attributes.” (Tang, col. 5, lines 20-23)
“constructing an input vector having the subset of input vector dimensions;” — Tang teaches this limitation. Tang discloses:
“the in-bag tuples can be visualized as datapoints plotted onto an n-dimensional space, where n equals the number of attributes associated with each tuple.” (Tang, col. 5, lines 18-20)
Tang further discloses:
“If a dimension is removed, the datapoints can be said to be projected into the subspace comprising the remaining attributes.” (Tang, col. 5, lines 20-23)
Tang also discloses that the random subspace may be selected by:
“randomly selecting attributes to remove from the subspace or randomly selecting the attributes that are included in the subspace.” (Tang, col. 5, lines 24-26)
and that:
“an original matrix associated with the training dataset can be reduced into a granule.”(Tang, col. 5, lines 30-32)
Thus, Tang teaches constructing reduced-dimensional input data/vectors having only the selected subset of attributes/dimensions.
“processing the input vector having the subset of input vector dimensions — Tang and Thomas teach this limitation and/or portions thereof.
Tang teaches processing a reduced-dimensional input vector. Tang discloses:
“The prediction module 230 can compare datapoints associated with the features against each of the hyperplane classifiers derived from the granules to derive granule predictions associated with the respective hyperplane classifiers.” (Tang, col. 6, lines 9-12)
Tang further discloses that the prediction module may:
“plot the unclassified new tuple onto a random subspace associated with a first granule and associated hyperplane classifier.” (Tang, col. 6, lines 14-16)
Neither Tang nor Baker teach these limitations and/or portions of:
“… and a stage knowledge map associated with an identification stage;”
“… through a stage knowledge map.”
Thomas, however, teaches these remaining limitations and/or portions of:
“… and a stage knowledge map associated with an identification stage;” – Thomas teaches retrieving a knowledge map/partition associated with a recognition stage. Thomas discloses:
“a pattern recognition engine maintains a knowledge element array which is a memory space for one or more knowledge maps.” (Thomas, col. 21, lines 30-34)
Thomas further discloses:
“Each knowledge map includes one or more knowledge elements, which itself includes a vector, and a category identifier.” (Thomas, col. 21, lines 32-34)
Thomas also discloses:
“A pattern recognition application can invoke a particular partition by identifying the partition when passing a learn, configure, or recognize command to the knowledge element array.” (Thomas, col. 21, lines 48-51)
Thomas further discloses:
“one recognition stage could use a first partition to provide a result.” (Thomas, col. 25, lines 49-50)
and:
“a subsequent recognition stage, using the same or a different input vector, could be performed in a different partition based on the opaque user data returned by the first recognition stage.” (Thomas, col. 25, lines 49-53)
Thus, Tang teaches selecting/retrieving the subset of vector dimensions used for reduced-dimensional processing, and Thomas teaches invoking/retrieving a particular partition/knowledge map associated with a recognition stage. The knowledge map associated with the particular recognition stage corresponds to the claimed stage knowledge map associated with an identification stage.
“… through a stage knowledge map.” – Thomas teaches processing an input vector through a knowledge map. Thomas discloses:
“A pattern recognition application can invoke a particular partition by identifying the partition when passing a learn, configure, or recognize command to the knowledge element array.” (Thomas, col. 21, lines 48-51)
Thomas further discloses:
“one recognition stage could use a first partition to provide a result.” (Thomas, col. 25, lines 49-50)
and:
“a subsequent recognition stage, using the same or a different input vector, could be performed in a different partition based on the opaque user data returned by the first recognition stage.” (Thomas, col. 25, lines 50-53)
Thomas also discloses:
“apply the input vector to the particular knowledge map for a recognition operation or a learning operation using the comparison technique.” (Thomas, col. 9, lines 58-60)
Thus, Tang teaches processing an input vector having a selected subset of dimensions, and Thomas teaches applying an input vector to a particular knowledge map/partition for a recognition operation in a staged recognition process. The particular knowledge map/partition used in the recognition stage corresponds to the claimed stage knowledge map.
Because claim 5 recites “one or more of” the listed operations, Tang’s teaching of constructing an input vector having the subset of input vector dimensions is sufficient to meet the additional limitation of claim 5. Additionally, Thomas teaches the knowledge-map/stage aspects of the other listed alternatives.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to further implement the Baker/Tang/Jones system using Thomas’s knowledge-map/partition structure. Baker teaches staged machine-learning classifiers, Tang teaches reduced-dimensional input/training data produced by selecting/removing attributes, and Jones teaches staged rejection/retention processing. Thomas teaches a known pattern-recognition implementation in which knowledge maps include knowledge elements having vectors and category identifiers, and in which recognition operations are organized as configured partitions arranged in serial or hierarchical stages. A person of ordinary skill in the art would have been motivated to use Thomas’s knowledge-map partition structure in the Baker/Tang/Jones staged classifier system to maintain the vector/category knowledge used by each stage and to apply the appropriate vector configuration and knowledge map during staged recognition operations.
Regarding claim 7, Baker, in view of Tang, further in view of Jones, and further in view of Thomas, teaches, the multistage classification system of claim 5, the logic operable to cause the one or more processors to:
“determine whether — Baker teaches this limitation at least in part. Baker discloses that:
“The multi-stage system includes a network of machine learning systems arranged into a series of sequential stages, wherein a prior stage feeds data into a subsequent stage until a final stage is reached.” (Baker, pg. 2, ¶[0018])
Baker further discloses:
“each non-final stage machine learning system functions as a classifier selecting which of the machine learning systems in the next stage is to receive each item of data.” (Baker, pg. 2, ¶[0019])
Thus, Baker teaches determining stage progression through related sequential stages because each non-final stage selects which next-stage system receives the data until the final stage is reached.
“continue to iterate through stages in response to determining that not all of the — Baker teaches this limitation at least in part. Baker discloses that:
“The multi-stage system includes a network of machine learning systems arranged into a series of sequential stages, wherein a prior stage feeds data into a subsequent stage until a final stage is reached.” (Baker, pg. 2, ¶[0018])
Baker further discloses that:
“The final classification decision is made by the one or more classifiers 104a-e in the final stage 104 that are selected by the non-final stages 101, 102.” (Baker, pg. 2, ¶[0019])
Thus, Baker teaches continuing through stages until the final stage is reached and a final classification decision is made.
Baker does not expressly teach these limitations and/or portions of:
“determine whether “
“continue to iterate through stages in response to determining that not all of the
“in response to determining that all of the related stages have been exercised, determine a probability of the input vector belonging to one or more classes based on neighboring classes.”
Thomas, however, teaches these limitations and/or the portion(s) of:
“determine whether “ – Thomas discloses:
“An overall pattern recognition process may be defined or configured as a series or set of individual pattern recognition operations, each associated with a configured partition.” (Thomas, col. 21, lines 59-61)
Thomas further discloses that partitions can be arranged:
“in a serial or hierarchical relationship” (Thomas, col. 21, line 64)
Thomas also discloses:
“one recognition stage could use a first partition to provide a result.” (Thomas, col. 25, lines 49-50)
and:
“a subsequent recognition stage, using the same or a different input vector, could be performed in a different partition based on the opaque user data returned by the first recognition stage.” (Thomas, col. 25, lines 49-53)
Thus, Thomas teaches determining whether stages related by a serial or hierarchical recognition relationship have been exercised because the subsequent recognition stage is performed based on the result returned by the first recognition stage.
“continue to iterate through stages in response to determining that not all of the To the extent Baker does not expressly teach continuing to iterate through all “related stages,” Thomas teaches this limitation. Thomas discloses:
“a subsequent recognition stage, using the same or a different input vector, could be performed in a different partition based on the opaque user data returned by the first recognition stage. This can continue for several levels.” (Thomas, col. 25, lines 49-54)
Thomas further discloses:
“Through experimentation, the correct numbers of levels are determined along with what to train/recognize in each level and what to feed up to higher levels.” (Thomas, col. 26, lines 55-58)
Thus, Thomas teaches continuing recognition through multiple related stages/levels until the appropriate levels have been exercised.
“in response to determining that all of the related stages have been exercised, determine a probability of the input vector belonging to one or more classes based on neighboring classes.” — Thomas teaches this limitation, to the extent the limitation can be understood in view of the §112(b) rejection above. Thomas discloses that:
“In this fashion probabilistic results can be feed further into the hierarchy.” (Thomas, col. 26, lines 50-51)
Thomas further discloses:
“An example would be the lower level results are that there is an 80% probability, as opposed to a binary result in the simpler hierarchy.” (Thomas, col. 26, lines 51-54)
Thomas also discloses recognition results based on categories and neighboring/multiple matching knowledge elements. Thomas discloses:
“Indeterminate Recognition (706)—The test vector fell within the current influence fields of more than one knowledge element and those knowledge elements were of different categories.” (Thomas, col. 28, lines 17-20)
Thomas further discloses that, in that case:
“the category the smallest distance away can be used, the majority category value of the knowledge elements matched can be used, or as with the Not Recognized state, additional training may be warranted.” (Thomas, col. 28, lines 20-24)
Thus, Thomas teaches determining a probability/classification based on neighboring or multiple matched classes because Thomas determines recognition using knowledge elements of different categories within influence fields and may use the closest category or majority category value of matched knowledge elements. Thomas further teaches probabilistic results in a hierarchy. Therefore, Thomas teaches determining a probability of the input vector belonging to one or more classes based on neighboring classes.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to further implement the Baker/Tang/Jones system using Thomas’s serial or hierarchical knowledge-map recognition stages and probabilistic recognition results. Baker teaches staged machine-learning classification, Tang teaches reduced-dimensional vector processing, and Jones teaches staged rejection/retention processing. Thomas teaches that pattern recognition operations may be arranged in serial or hierarchical relationships, that subsequent recognition stages may be performed based on results returned by prior stages, and that probabilistic results may be fed through the hierarchy. A person of ordinary skill in the art would have been motivated to use Thomas’s serial/hierarchical recognition structure and probability-based category determination in the Baker/Tang/Jones multistage classifier system to continue recognition through related stages and provide a probabilistic classification result when the related stages have been exercised.
Regarding claim 8, Baker, in view of Tang, further in view of Jones, and further in view of Thomas, teaches the multistage classification system of claim 7,
“wherein the corresponding action comprises one or more of generating an alert in response to determining that the input vector belongs to a class” – Baker does not teach this limitation. Jones, however, teaches this limitation at least in part. Jones discloses that after retained subwindows are processed, the system may perform:
“output processing such as highlighting the instances 38 of objects in an output image 34 for viewing by an end-user.” (Jones, pg. 6, ¶[0067])
Thus, Jones teaches generating a visible output/notification to a user in response to determining that an input image subwindow belongs to an object class. Highlighting detected object instances in an output image corresponds to generating an alert or notification responsive to determining that the input belongs to a class.
To the extent the probability alternative is considered, neither Jones nor Baker expressly teaches this limitation:
“or determining that the input vector belongs to the class with a probability of substantially one.”
Thomas, however, teaches this remaining limitation:
“or determining that the input vector belongs to the class with a probability of substantially one.” — Thomas teaches this limitation, to the extent the limitation can be understood in view of the §112(b) rejection above. Thomas discloses probability-based recognition in a hierarchical recognition process:
“In this fashion probabilistic results can be feed further into the hierarchy.” (Thomas, col. 26, lines 50-51)
Thomas further discloses:
“An example would be the lower level results are that there is an 80% probability, as opposed to a binary result in the simpler hierarchy.” (Thomas, col. 26, lines 51-54)
Thomas also discloses exact identification/recognition when an unknown vector falls within influence fields associated with a single category:
“If the unknown vector is within the influence field of knowledge elements within a single category, it is termed ‘exact identification’.” (Thomas, col. 16, lines 60-64)
Thomas further discloses:
“Exact Recognition (702)—The test vector fell within the current influence fields of one or more knowledge elements and all of those knowledge elements were in the same category.” (Thomas, col. 28, lines 9-12)
Thus, Thomas teaches determining that an input vector belongs to a class with exact recognition or probability-based recognition. To the extent “probability of substantially one” is understood as an effectively certain or exact classification, Thomas’s exact-recognition teaching corresponds to determining that the input vector belongs to the class with a probability of substantially one.
Because claim 8 recites “one or more of” the listed actions, Jones’s teaching of generating an output/alert by highlighting detected object instances is sufficient to meet the additional limitation of claim 8. Additionally, Thomas teaches the probability/exact-recognition alternative.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to configure the Baker/Tang/Jones/Thomas system to generate an alert or output when an input is classified, as taught by Jones. Baker teaches classification to a classification output, Tang teaches reduced-dimensional classifier processing, Jones teaches output processing by highlighting detected object instances for an end user, and Thomas teaches exact/probabilistic recognition. A person of ordinary skill in the art would have been motivated to provide an alert or output indicating the classification result so that the result of the multistage classification process is communicated to a user or downstream system.
Regarding claims 10, 13, 15, 18, 21, and 23
Claims 10, 13, and 15 are the non-transitory computer-readable medium counterparts of claims 2, 5, and 7, respectively. Claims 18, 21, and 23 are the method counterparts of claims 2, 5, and 7, respectively.
Each of claims 10, 13, and 15 recites a non-transitory computer-readable medium storing instructions to cause one or more processors to perform substantially the same functions recited in corresponding system claims 2, 5, and 7. Each of claims 18, 21, and 23 recites computer-implemented method steps corresponding to the functions recited in corresponding system claims 2, 5, and 7.
Baker, Tang, and Jones teach the base multistage classification system and the corresponding reduced-dimensional staged-classification processing for the reasons set forth above with respect to claim 1.
Tang teaches identifying a subset of input vector dimensions and constructing reduced-dimensional input/training vectors by selecting or removing attributes/dimensions, as set forth above with respect to claims 2 and 5.
Thomas teaches knowledge maps, knowledge elements including vectors and category identifiers, configured partitions associated with pattern-recognition operations, serial or hierarchical recognition stages, and probabilistic/exact recognition, as set forth with respect to claims 2, 5, and 7.
Expressing the same computer-implemented operations as stored instructions on a non-transitory computer-readable medium or as method steps does not, without more, impart patentable distinction over the previously established system combination. See MPEP §§2113 and 2114.
Accordingly, claims 10, 13, 15, 18, 21, and 23 are obvious over Baker in view of Tang, further in view of Jones, and further in view of Thomas for the same reasons discussed for corresponding system claims 2, 5, and 7.
Conclusion
THIS ACTION IS MADE FINAL. 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 Paul Coleman whose telephone number is (571)272-4687. The examiner can
normally be reached Mon-Fri.
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,
David Yi can be reached at (571) 270-7519. 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.
/PAUL COLEMAN/
Examiner, Art Unit 2126
/DAVID YI/Supervisory Patent Examiner, Art Unit 2126
Prosecution Insights