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 .
Priority
Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d), based on an application filed in FEDERAL REPUBLIC OF GERMANY on 11/11/2021. Receipt is acknowledged of certified copies of papers required by 37 CFR 1.55.
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 07/06/2026 has been entered. Claim(s) 1-14 are pending. Claim(s) 2, 5, 8, 11 have been withdrawn from consideration.
Response to Amendment
The amendment filed on July 6th, 2026 has been entered and Claims 1-14 are pending. Claims 2, 5, 8 and 11 have been withdrawn from consideration. Applicant’s amendments to the Claim(s) have overcome every 101 rejection(s) previously set forth in the Final Office Action mailed on April 6th, 2026.
Response to Arguments
Applicant's arguments filed July 6th, 2026 have been fully considered but they are not persuasive.
Applicant’s arguments with respect to U.S.C 101 rejection of a clear improvement. This is not persuasive. The claim 1 language broadly recites “training the machine learning algorithm to control at least one function of a controllable system” and it does not identify a particular controllable system, a particular function of the system being controlled, or any other manner in which the trained machine learning algorithm controls the function. Applicant relies on the specification to state the technology improvement, but the claim must itself reflect the disclosed improvement or steps that provide that improvement. See (MPEP 2106.05(a), To show that the involvement of a computer assists in improving the technology, the claims must recite the details regarding how a computer aids the method, the extent to which the computer aids the method, or the significance of a computer to the performance of the method. Merely adding generic computer components to perform the method is not sufficient. Thus, the claim must include more than mere instructions to perform the method on a generic component or machinery to qualify as an improvement to an existing technology.) See updated 101 rejection.
Applicant’s arguments with respect to DeVries does not anticipate “applying robust statistics…” This is not persuasive. Examiner does not rely on DeVries to teach or suggest the limitation. Instead, the examiner noted in the previous office action the limitation is taught or suggest to DeVries in view of Atkeson. See updated rejection.
Applicant’s arguments with respect to “combination with Atkeson does not…” This is not persuasive. The office action rely on Atkeson to teach or suggest the applying robust statistics for detecting and remove outliers used in combination with DeVries that is relied on to teach or suggest the approximate nearest neighbors. The Examiner relies on the broadest reasonable interpretation (BRI) of the claim language where robust statistics is applied in order to detect outliers which the applicant relies on narrow interpretation of the claim. See updated rejection.
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.
Claim 1, 3-4, 6-7, 9-10, 12-14 is/are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., an abstract idea) without significantly more.
Regarding claim 1:
Step 1 (whether a claim is to a statutory category):
Yes, the claim is within the four statutory categories (a process, machine, manufacture or composition of matter). Claim 1 recites a method, therefore, falls within a process category.
Step 2A Prong 1 (whether a claim is directed to a judicial exception):
Yes, “identifying, based on the additional input data point, approximate nearest neighbors of the additional input data point from the input data points of the first training data;” describes a mental process (i.e., evaluation, judgement) wherein approximating the distance of the additional input data point to the first training data’s input data point recites concepts that can be practically performed in the human mind and/or with physical aid (i.e., pen and paper, and/or calculator) (see MPEP 2106.04(a)(2)(III)).) And, “applying robust statistics to the output data values associated with the approximate nearest neighbors of the additional input data point, in order to detect outliers in the output data values associated with the approximate nearest neighbors of the additional input data point, and” wherein applying statistics recites a mathematical concept, as it involves mathematical relationships, mathematical formulas or equations, mathematical calculations (see MPEP 2106.04(a)(2), (I)), and a mental process (i.e., observation, evaluation) wherein observing the data values to evaluate outliers recites concepts that can be practically performed in the human mind and/or with physical aid (i.e., pen and paper) (see MPEP 2106.04(a)(2)(III)).) And “determining an additional output data value associated with the additional input data point based on the output data values that are associated with the approximate nearest neighbors of the additional input data point and that do not represent an outlier;” describes a mental process (i.e., evaluation, judgement) wherein evaluating the data values and judging which values are neighbors to associate the data values to the additional data point recites concepts that can be practically performed in the human mind and/or with physical aid (i.e., pen and paper) (see MPEP 2106.04(a)(2)(III)).)
Step 2A Prong 2 (evaluate whether the claim recites additional elements that integrate
the exception into a practical application):
No, “generating training data for training a machine learning algorithm,” and “receiving first training data for training the machine learning algorithm, the first training data comprising a plurality of data pairs that each include an input data point and an output value associated with the input data point;” the claim does not recite additional elements that integrate the judicial exception into a practical application with the words "apply it" (or an equivalent), such as mere (i.e., selecting a particular data source or type of data to be manipulated) to implement an abstract idea on a computer (see MPEP 2106.05(f)). and “receiving an additional input data point;” the claim does not recite additional elements that integrate the judicial exception into a practical application with the words "apply it" (or an equivalent), such as mere (i.e., selecting a particular data source or type of data to be manipulated) to implement an abstract idea on a computer (see MPEP 2106.05(f)). and “forming additional training data having an additional data pair including the additional input data point and the additional output data value associated with the additional data input point; and the claim does not recite additional elements that integrate the judicial exception into a practical application with the words "apply it" (or an equivalent), such as mere (i.e., selecting a particular data source or type of data to be manipulated) to implement an abstract idea on a computer (see MPEP 2106.05(f)). And, training the machine learning algorithm to control at least one function of a controllable system, the machine learning algorithm being trained based on the first training data and the additional training data” the limitation is an additional element that generally links the use of the judicial exception to a particular technological environment or field of use. It broadly recites a controllable system without identifying the system, the function being controlled, or how the trained machine learning algorithm controls the function. See MPEP § 2106.05(h). And, the claim does not recite additional elements that integrate the judicial exception into a practical application with the words "apply it" (or an equivalent), to implement an abstract idea on a computer (see MPEP 2106.05(f)).
Step 2B (Inventive concept):
No, it does not add significantly more since the intended practical application is well-understood, routine, and conventional and stated at a generic level (i.e. “apply it”, see MPEP 2106.05(f)).
Therefore, claim 1 is ineligible.
Regarding claim 3:
Further modifies the abstract idea of claim 1.
Step 2A Prong 1 (whether a claim is directed to a judicial exception):
Yes, “determining a median of the output data values associated with the approximate nearest neighbors of the additional input data point” describes a mental process wherein evaluating data values and making a judgement based on a median of the associated approximate nearest neighbors of the additional data point recites concepts that can be practically performed in the human mind and/or with physical aid (i.e., pen and paper) (see MPEP 2106.04(a)(2)(III)).)
Step 2A Prong 2 (evaluate whether the claim recites additional elements that integrate
the exception into a practical application):
No, “wherein determining the additional output data value associated with the additional input data point from output data values associated with the approximate nearest neighbors of the additional input data point comprises” the claim does not recite additional elements that integrate the judicial exception into a practical application with the words "apply it" (or an equivalent), such as mere (i.e., selecting a particular data source or type of data to be manipulated) to implement an abstract idea on a computer (see MPEP 2106.05(f)).
Step 2B (Inventive concept):
No, it does not add significantly more since the intended practical application is well-understood, routine, and conventional and stated at a generic level (i.e. “apply it”, see MPEP 2106.05(f)).
Therefore, claim 3 is ineligible.
Regarding claim 4:
Further modifies the abstract idea of claim 1.
Step 2A Prong 2 (evaluate whether the claim recites additional elements that integrate
the exception into a practical application):
No, “wherein the first training data comprise sensor data” does not recite additional elements that integrate the judicial exception into a practical application with the words "apply it" (or an equivalent), such as mere (i.e., selecting a particular data source or type of data to be manipulated) to implement an abstract idea on a computer (see MPEP 2106.05(f)).
Step 2B (Inventive concept):
No, it does not add significantly more since the intended practical application is well-understood, routine, and conventional and stated at a generic level (i.e. “apply it”, see MPEP 2106.05(f)).
Therefore, claim 4 is ineligible.
Regarding claim 6:
Further modifies the abstract idea of claim 1.
Step 2A Prong 2 (evaluate whether the claim recites additional elements that integrate
the exception into a practical application):
No, “providing the machine learning algorithm for controlling the at least one function of the controllable system, the machine learning algorithm having been trained according to the method of Claim 1;” and “and controlling the at least one function of the controllable system based on the trained machine learning algorithm.” describes an additional element as “apply it”, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(f)).
Step 2B (Inventive concept):
No, it does not add significantly more since the intended practical application is well-understood, routine, and conventional and stated at a generic level (i.e. “apply it”, see MPEP 2106.05(f)).
Therefore, claim 6 is ineligible.
Regarding claim 7:
Step 1 (whether a claim is to a statutory category):
Yes, the claim is within the four statutory categories (a process, machine, manufacture or composition of matter). Claim 7 recites a device, therefore, falls within a machine category.
Step 2A Prong 1 (whether a claim is directed to a judicial exception):
Yes, “identify, based on the additional input data point, approximate nearest neighbors of the additional input data point from the input data points of the first training data;” describes a mental process (i.e., evaluation, judgement) wherein approximating the distance of the additional input data point to the first training data’s input data point recites concepts that can be practically performed in the human mind and/or with physical aid (i.e., pen and paper, and/or calculator) (see MPEP 2106.04(a)(2)(III)).) And, “apply robust statistics to the output data values associated with the approximate nearest neighbors of the additional input data point, in order to detect outliers in the output data values associated with the approximate nearest neighbors of the additional input data point, and” wherein applying statistics recites a mathematical concept, as it involves mathematical relationships, mathematical formulas or equations, mathematical calculations (see MPEP 2106.04(a)(2), (I)), and a mental process (i.e., observation, evaluation) wherein observing the data values to evaluate outliers recites concepts that can be practically performed in the human mind and/or with physical aid (i.e., pen and paper) (see MPEP 2106.04(a)(2)(III)).) And “determine, an additional output data value associated with the additional input data point based on the output data values that are associated with the approximate nearest neighbors of the additional input data point; and that do not represent an outlier” describes a mental process (i.e., evaluation, judgement) wherein evaluating the data values and judging which values are neighbors to associate the data values to the additional data point recites concepts that can be practically performed in the human mind and/or with physical aid (i.e., pen and paper) (see MPEP 2106.04(a)(2)(III)).)
Step 2A Prong 2 (evaluate whether the claim recites additional elements that integrate the exception into a practical application):
No, “forming additional training data having an additional data pair including the additional input data point and the additional output data value associated with the additional data input point; and” the claim does not recite additional elements that integrate the judicial exception into a practical application with the words "apply it" (or an equivalent), such as mere (i.e., selecting a particular data source or type of data to be manipulated) to implement an abstract idea on a computer (see MPEP 2106.05(f)). And, “a memory configured to store first training data, an additional input data point, and program code, the first training data comprising a plurality of data pairs that each include an input data point and an output data value associated with the input data point; and a processor operably connected to the memory and configured to execute the program code to:” the claim does not recite additional elements that integrate the judicial exception into a practical application with the words "apply it" (or an equivalent), such as mere (i.e., selecting a particular data source or type of data to be manipulated) to implement an abstract idea on a computer (see MPEP 2106.05(f)). And, train the machine learning algorithm to control at least one function of a controllable system, the machine learning algorithm being trained based on the first training data and the additional training data” the limitation is an additional element that generally links the use of the judicial exception to a particular technological environment or field of use. It broadly recites a controllable system without identifying the system, the function being controlled, or how the trained machine learning algorithm controls the function. See MPEP § 2106.05(h). And, the claim does not recite additional elements that integrate the judicial exception into a practical application with the words "apply it" (or an equivalent), to implement an abstract idea on a computer (see MPEP 2106.05(f)).
Step 2B (Inventive concept):
No, it does not add significantly more since the intended practical application is well-understood, routine, and conventional and stated at a generic level (i.e. “apply it”, see MPEP 2106.05(f)).
Therefore, claim 7 is ineligible.
Regarding claim 9:
Further modifies the abstract idea of claim 7.
Step 2A Prong 1 (whether a claim is directed to a judicial exception):
Yes, “determine the additional output associated with the additional input data point by determining a median of the output data values associated with the approximate nearest neighbors of the additional input data point” describes a mental process wherein evaluating data values and making a judgement based on a median of the associated approximate nearest neighbors of the additional data point recites concepts that can be practically performed in the human mind and/or with physical aid (i.e., pen and paper) (see MPEP 2106.04(a)(2)(III)).)
Step 2A Prong 2 (evaluate whether the claim recites additional elements that integrate the exception into a practical application):
No, the claim does not recite additional elements that integrate the judicial exception into a practical application with the words "apply it" (or an equivalent), such as mere (i.e., selecting a particular data source or type of data to be manipulated) to implement an abstract idea on a computer (see MPEP 2106.05(f)).
Step 2B (Inventive concept):
No, it does not add significantly more since the intended practical application is well-understood, routine, and conventional and stated at a generic level (i.e. “apply it”, see MPEP 2106.05(f)).
Therefore, claim 9 is ineligible.
Regarding claim 10:
Further modifies the abstract idea of claim 7.
Step 2A Prong 2 (evaluate whether the claim recites additional elements that integrate
the exception into a practical application):
No, “wherein the first training data comprise sensor data” does not recite additional elements that integrate the judicial exception into a practical application with the words "apply it" (or an equivalent), such as mere (i.e., selecting a particular data source or type of data to be manipulated) to implement an abstract idea on a computer (see MPEP 2106.05(f)).
Step 2B (Inventive concept):
No, it does not add significantly more since the intended practical application is well-understood, routine, and conventional and stated at a generic level (i.e. “apply it”, see MPEP 2106.05(f)).
Therefore, claim 10 is ineligible.
Regarding claim 12:
Further modifies the abstract idea of claim 7.
Step 2A Prong 2 (evaluate whether the claim recites additional elements that integrate
the exception into a practical application):
No, “a controller configured to control the at least one function of the controllable system based on the machine learning algorithm, the machine learning algorithm trained by the control device of Claim 7.” does not recite additional elements that integrate the judicial exception into a practical application with the words "apply it" (or an equivalent), such as mere (i.e., selecting a particular data source or type of data to be manipulated) to implement an abstract idea on a computer (see MPEP 2106.05(f)).
Step 2B (Inventive concept):
No, it does not add significantly more since the intended practical application is well-understood, routine, and conventional and stated at a generic level (i.e. “apply it”, see MPEP 2106.05(f)).
Therefore, claim 12 is ineligible.
Regarding claim 13 and analogous claim 14:
Further modifies the abstract idea of claim 1.
Step 2A Prong 2 (evaluate whether the claim recites additional elements that integrate the exception into a practical application):
No, “wherein the controllable system is one of (i) an injection system of an internal combustion engine, (ii) an analyzer configured to analyze samples for presence of viruses, (iii) a driver assistance system of a motor vehicle, (iv) a robotic system, (v) a kitchen appliance, or (vi) a washing machine” the limitation is an additional element that generally links the use of the judicial exception to a particular technological environment or field of use. It broadly recites a controllable system without identifying the system and/or lists more which makes it unclear what the system is, the function being controlled, or how the trained machine learning algorithm controls the function. See MPEP § 2106.05(h). And, the claim does not recite additional elements that integrate the judicial exception into a practical application with the words "apply it" (or an equivalent), to implement an abstract idea on a computer (see MPEP 2106.05(f)).
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.
Claim(s) 1, 3-4, 6-7, 9-10 and 12 is/are rejected under 35 U.S.C. 103 as being unpatentable over DeVries et al., Non-Patent Literature (“Dataset Augmentation in Feature Space”), in view of Atkeson et al., Non-Patent Literature (“Memory-based neural networks for robot learning”) and further in view of Gupta et al., Non-Patent Literature (“ProtoNN: Compressed and Accurate kNN for Resource-scarce Devices”)
Regarding claim 1:
DeVries teaches:
A method for generating training data for training a machine learning algorithm, (Introduction, paragraph 4, (Introduction, paragraph 1, “dataset augmentation, wherein the existing data is transformed in some way to create new data [generating training data] that appears to come from the same (conditional) data generating distribution”…(Introduction, paragraph 4, “We show that models trained on datasets that have been augmented using our technique outperform models trained only on data from the original dataset. (i.e., wherein the dataset augmentation is a technique used to generate data to train a model))”)
the first training data comprising a plurality of data pairs that each include an input data point and an output data value associated with the input data point; (Section 3.2, “where i indexes the elements of a context vector which corresponds to data points from the training set [training data]”…“ For each sample in the dataset, we find its K nearest neighbours in feature space which share its class label”)…“For each sample [input data point] in the dataset, we find its K nearest neighbours in feature space which share its class label [output data value]. For each pair [data pairs] of neighbouring context vectors, a new context vector can then be generated”)
receiving first training data for training the machine learning algorithm; (Introduction, paragraph 4, “We show that models trained on datasets that have been augmented using our technique outperform models trained only on data from the original dataset [first training data]. (i.e., wherein the ‘original dataset’ is interpreted as the first training data used to train the model.)”)
receiving an additional input data point; (“In our experiments, we use λ = 0.5 so that the new sample [additional data point] balances properties of both original samples (i.e., wherein providing the ‘new sample’ is interpreted as additional data point)”)
identifying, based on the additional input data point, approximate nearest neighbors of the additional input data point from the data points of the first training data; (Section 3.2, “For each sample in the dataset [data points of the first training data], we find its K nearest neighbours [nearest neighbors] in feature space which share its class label. For each pair of neighbouring context vectors, a new context vector can then be generated using interpolation: c 0 = (ck − cj )λ + cj (2) where c 0 is the synthetic context vector, ci and cj are neighbouring context vectors, (i.e., wherein finding the approximate nearest neighbors to the additional data point to the original data point )”)
determining an additional output data value associated with the additional input data point based on the output data values that are associated with the approximate nearest neighbors of the additional input data point; (“For each sample [data point] in the dataset, we find its K nearest neighbours in feature space which share its class label [data value]. For each pair of neighbouring context vectors, a new context vector can then be generated [additional data point] (i.e., wherein each pair of neighboring context vectors include its ‘class label’ which is interpreted to be a determined data value)”)
forming additional training data having an additional data pair including the additional input data point and the additional output data value associated with the additional input data point;(Section 3.2, “For each sample [data point] in the dataset, we find its K nearest neighbours in feature space which share its class label [data value]. For each pair [data pair] of neighbouring context vectors, a new context vector can then be generated [additional data point] (i.e., wherein ‘new context vector’ is interpreted as forming additional training data having a data pair of the data point and data value)”)
Atkeson teaches:
applying robust statistics to the output data values associated with the approximate nearest neighbors of the additional input data point, in order to detect outliers in the output data values associated with the approximate nearest neighbors of the additional input data point; (Introduction, “An important problem in motor learning is approximating a continuous function from samples of the function’s inputs and outputs [data value]. This paper explores a neural network architecture that simply remembers experiences (samples) and builds a local model to answer any particular query [additional data point] (an input for which the function’s output is desired [data value])”…(Section 3, paragraph 3, “Often, a group of similar experiences, or nearest neighbors [nearest neighbors], is used to form a local model, and then that model is used to predict the desired value for a new point (i.e., wherein the ‘output’ is interpreted as the data value that is associated to the ‘query’ (additional data point) using nearest neighbors)”…(Section 9, “Linear regression is not robust with respect to outliers. This also holds for locally weighted regression, although the influence of outliers will not be noticed unless the outliers lie close enough to a query point. In Fig. 6(a) we added three outliers to the test data of Fig. 5 to demonstrate this effect; the plots in Fig. 6 should be compared to Fig. 5(c). [471 applied the median absolute deviation procedure from robust statistics [robust statistics] [32] to globally remove outliers”)
and that do not represent an outlier; (Section 9, “Linear regression is not robust with respect to outliers. This also holds for locally weighted regression, although the influence of outliers will not be noticed (i.e., wherein the ‘influence of outliers will not be noticed’ is interpreted as the data value is not determined by the outliers)
A person of ordinary skill in the art would reasonably find the teachings of Atkeson to be helpful in solving the problem of handling outliers using robust statistics present in DeVries. In view of the teachings of Atkeson it would have been obvious for a person of ordinary skill in the art to apply the teachings of Atkeson to DeVries before the effective filing date of the claimed invention in order to improve the efficiency of computer resources (Atkeson, Abstract, “In this paper their nearest neighbor network is augmented with a local model network, which fits a local model to a set of nearest neighbors. This network design is equivalent to a statistical approach known as locally weighted regression, in which a local model is formed to answer each query, using a weighted regression in which nearby points (similar experiences) are weighted more than distant points (less relevant experiences). We illustrate this approach by describing how it has been used to enable a robot to learn a difficult juggling task.”)
Gupta teaches:
train the machine learning algorithm to control at least one function of the controllable system, the machine learning algorithm being trained based on the first training data and the additional training data (Introduction, paragraph 1, “IoT devices have the potential to provide realtime, local, sensor-based solutions for a variety of areas like housing, factories, farming, even everyday utilities like toothbrushes and spoons. The ability to use machine learning on data collected from IoT sensors opens up a myriad of possibilities. For example, smart factories measure temperature, noise and various other parameters of their machines. ML based anomaly detection models [machine learning algorithms] can then be applied on this sensor data to preemptively schedule maintenance of a machine [controllable system] and avoid failure (i.e., wherein the ‘machine’ is interpreted as being the controllable system and the controlling function is interpreted as being ‘able to schedule maintenance of the machine’, hence that ‘ML based anomaly detection models can be applied to the data’ is interpreted as the controlling function is done based on the training)”)
A person of ordinary skill in the art would reasonably find the teachings of Gupta to be helpful in solving the problem of handling data for prediction modeling present in DeVries. In view of the teachings of Gupta it would have been obvious for a person of ordinary skill in the art to apply the teachings of Gupta to DeVries before the effective filing date of the claimed invention in order to improve the efficiency of computer resources (Gupta, Abstract, “ProtoNN models can be deployed even on devices with puny storage and computational power (e.g. an Arduino UNO with 2kB RAM) to get excellent prediction accuracy. ProtoNN derives its strength from three key ideas: a) learning a small number of prototypes to represent the entire training set, b) sparse low dimensional projection of data, c) joint discriminative learning of the projection and prototypes with explicit model size constraint. We conduct systematic empirical evaluation of ProtoNN on a variety of supervised learning tasks (binary, multi-class, multi-label classification) and show that it gives nearly state-of-the-art prediction accuracy on resource-scarce devices while consuming several orders lower storage, and using minimal working memory.”
Regarding claim 3 and analogous claim 9:
DeVries, as modified by Atkeson and Gupta, teaches the method of claim 1.
DeVries does not explicitly teach:
wherein determining the additional output data value associated with the additional input data point from output data values associated with the approximate nearest neighbors of the additional input data point comprises
determining a median of the output data values associated with the approximate nearest neighbors of the additional input data point.
Atkeson further teaches:
wherein determining the data value associated with the additional data point from data values associated with the nearest neighbors of the additional data point comprises (Introduction, “An important problem in motor learning is approximating a continuous function from samples of the function’s inputs and outputs [data value]. This paper explores a neural network architecture that simply remembers experiences (samples) and builds a local model to answer any particular query [additional data point] (an input for which the function’s output is desired [data value])”…(Section 3, paragraph 3,“Often, a group of similar experiences, or nearest neighbors [nearest neighbors], is used to form a local model, and then that model is used to predict the desired value for a new point (i.e., wherein the ‘output’ is interpreted as the data value that is associated to the ‘query’ (additional data point) using nearest neighbors)”
Gupta teaches:
determining a median of the data values associated with the nearest neighbors of the additional data point (Section 4, paragraph 5, “we run k-means [nearest neighbors] clustering in the transformed space on data points belonging to each class [data value] and pick the cluster centers [determining a median] as our prototypes [additional data point]. We use this approach for binary and multi-class problems (i.e., wherein the data value of the additional data point is the ‘center’ which is interpreted as the median)”)
A person of ordinary skill in the art would reasonably find the teachings of Gupta to be helpful in solving the problem of handling data for prediction modeling present in DeVries. In view of the teachings of Gupta it would have been obvious for a person of ordinary skill in the art to apply the teachings of Gupta to DeVries before the effective filing date of the claimed invention in order to improve the efficiency of computer resources (Gupta, Abstract, “ProtoNN models can be deployed even on devices with puny storage and computational power (e.g. an Arduino UNO with 2kB RAM) to get excellent prediction accuracy. ProtoNN derives its strength from three key ideas: a) learning a small number of prototypes to represent the entire training set, b) sparse low dimensional projection of data, c) joint discriminative learning of the projection and prototypes with explicit model size constraint. We conduct systematic empirical evaluation of ProtoNN on a variety of supervised learning tasks (binary, multi-class, multi-label classification) and show that it gives nearly state-of-the-art prediction accuracy on resource-scarce devices while consuming several orders lower storage, and using minimal working memory.”
Regarding claim 4 and analogous claim 10:
DeVries, as modified by Atkeson and Gupta, teaches the method of claim 1.
DeVries further teaches:
wherein the first training data comprise sensor data (Conclusion, paragraph 1, “We demonstrate our technique quantitatively on five datasets from different domains (speech, sensor processing, motion capture, and images) [sensor data] (i.e., wherein the sensor data can be data captured by a sensor (cited in spec, [0024]) {hence, ‘sensor processing’ and ‘motion capture’ is interpreted as generating sensor data})”)
Regarding claim 6 and analogous claim 12:
DeVries, as modified by Atkeson and Gupta, teaches the method of claim 5.
DeVries further teaches:
the machine learning algorithm having been trained according to the method of Claim 5; and claim 5: “providing first training data” (Introduction, paragraph 4, “We show that models trained on datasets that have been augmented using our technique outperform models trained only on data from the original dataset [first training data]. (i.e., wherein the ‘original dataset’ is interpreted as the first training data used to train the model.)”) and “forming additional training data according to the method of Claim 1; and (Section 3.2, “For each sample [data point] in the dataset, we find its K nearest neighbours in feature space which share its class label [data value]. For each pair [data pair] of neighbouring context vectors, a new context vector can then be generated [additional data point] (i.e., wherein ‘new context vector’ is interpreted as forming additional training data having a data pair of the data point and data value)”) and “training the machine learning algorithm based on the first training data and the additional training data” (Section 4.3, paragraph 1-2, “For our first quantitative test we use the Arabic Digits dataset (Lichman, 2013) which contains 8,800 samples of time series mel-frequency cepstrum coefficients (MFCCs) extracted from audio clips of spoken Arabic digits [first training data]. Thirteen MFCCs are available for each time step in this dataset. To preprocess the data we apply global normalization. To evaluate our data augmentation techniques we used the official train/test split and trained ten models with different random weight initializations. As a baseline model we trained a simple two layer MLP on the context vectors produced by a SA [additional training data]. Both models used 256 hidden units in each hidden layer. The MLP applied dropout with p = 0.5 after each dense layer. To evaluate the usefulness of different data augmentation techniques we trained a new baseline model on datasets that had been augmented with newly created samples. (i.e., wherein the additional training data and the first training data is used to improve training)”)
DeVries and Atkeson does not explicitly teach:
providing the machine learning algorithm for controlling the at least one function of the controllable system
controlling the at least one function of the controllable system based on the trained machine learning algorithm.
Gupta further teaches:
providing the machine learning algorithm for controlling the at least one function of the controllable system (Introduction, paragraph 1, “IoT devices have the potential to provide realtime, local, sensor-based solutions for a variety of areas like housing, factories, farming, even everyday utilities like toothbrushes and spoons. The ability to use machine learning on data collected from IoT sensors opens up a myriad of possibilities. For example, smart factories measure temperature, noise and various other parameters of their machines. ML based anomaly detection models [machine learning algorithms] can then be applied on this sensor data to preemptively schedule maintenance of a machine [controllable system] and avoid failure (i.e., wherein the ‘machine’ is interpreted as being the controllable system and the controlling function is interpreted as being able to schedule maintenance of the machine)”)
controlling the at least one function of the controllable system based on the trained machine learning algorithm (Introduction, paragraph 1, “IoT devices have the potential to provide realtime, local, sensor-based solutions for a variety of areas like housing, factories, farming, even everyday utilities like toothbrushes and spoons. The ability to use machine learning on data collected from IoT sensors opens up a myriad of possibilities. For example, smart factories measure temperature, noise and various other parameters of their machines. ML based anomaly detection models [machine learning algorithms] can then be applied on this sensor data to preemptively schedule maintenance of a machine [controllable system] and avoid failure (i.e., wherein the ‘machine’ is interpreted as being the controllable system and the controlling function is interpreted as being ‘able to schedule maintenance of the machine’, hence that ‘ML based anomaly detection models can be applied to the data’ is interpreted as the controlling function is done based on the training)”)
The motivation for claim 6 and 12 is the same motivation for claim 3.
Regarding claim 7:
DeVries further teaches:
the training data respectively comprise a data point and a data value associated with the data point, the control device comprising: (Section 3.2, “where i indexes the elements of a context vector which corresponds to data points [data point] from the training set [training data]”…“ For each sample in the dataset, we find its K nearest neighbours in feature space which share its class label [data value]”)
identify, based on the additional input data point, approximate nearest neighbors of the additional input data point from the input data points of the first training data; (Section 3.2, “For each sample in the dataset [data points of the first training data], we find its K nearest neighbours [nearest neighbors] in feature space which share its class label. For each pair of neighbouring context vectors, a new context vector can then be generated using interpolation: c 0 = (ck − cj )λ + cj (2) where c 0 is the synthetic context vector, ci and cj are neighbouring context vectors, (i.e., wherein finding the nearest neighbors to the additional data point to the original data point )”)
determine an additional output data value associated with the additional input data point based on output data values associated with the approximate nearest neighbors of the additional input data point, (“For each sample [data point] in the dataset, we find its K nearest neighbours in feature space which share its class label [output data value]. For each pair of neighbouring context vectors, a new context vector can then be generated [additional input data point] (i.e., wherein each pair of neighboring context vectors include its ‘class label’ which is interpreted to be a determined data value)”)
form additional training data having additional data pair including the additional input data point and the additional output data value associated with the additional input (Section 3.2, “For each sample [data point] in the dataset, we find its K nearest neighbours in feature space which share its class label [data value]. For each pair [data pair] of neighbouring context vectors, a new context vector can then be generated [additional data point]”)
Atkeson teaches:
applying robust statistics to the output data values associated with the approximate nearest neighbors of the additional input data point, in order to detect outliers in the output data values associated with the approximate nearest neighbors of the additional input data point; (Introduction, “An important problem in motor learning is approximating a continuous function from samples of the function’s inputs and outputs [data value]. This paper explores a neural network architecture that simply remembers experiences (samples) and builds a local model to answer any particular query [additional data point] (an input for which the function’s output is desired [data value])”…(Section 3, paragraph 3, “Often, a group of similar experiences, or nearest neighbors [nearest neighbors], is used to form a local model, and then that model is used to predict the desired value for a new point (i.e., wherein the ‘output’ is interpreted as the data value that is associated to the ‘query’ (additional data point) using nearest neighbors)”…(Section 9, “Linear regression is not robust with respect to outliers. This also holds for locally weighted regression, although the influence of outliers will not be noticed unless the outliers lie close enough to a query point. In Fig. 6(a) we added three outliers to the test data of Fig. 5 to demonstrate this effect; the plots in Fig. 6 should be compared to Fig. 5(c). [471 applied the median absolute deviation procedure from robust statistics [robust statistics] [32] to globally remove outliers”)
and that do not represent an outlier; (Section 9, “Linear regression is not robust with respect to outliers. This also holds for locally weighted regression, although the influence of outliers will not be noticed (i.e., wherein the ‘influence of outliers will not be noticed’ is interpreted as the data value is not determined by the outliers)
A person of ordinary skill in the art would reasonably find the teachings of Atkeson to be helpful in solving the problem of handling outliers using robust statistics present in DeVries. In view of the teachings of Atkeson it would have been obvious for a person of ordinary skill in the art to apply the teachings of Atkeson to DeVries before the effective filing date of the claimed invention in order to improve the efficiency of computer resources (Atkeson, Abstract, “In this paper their nearest neighbor network is augmented with a local model network, which fits a local model to a set of nearest neighbors. This network design is equivalent to a statistical approach known as locally weighted regression, in which a local model is formed to answer each query, using a weighted regression in which nearby points (similar experiences) are weighted more than distant points (less relevant experiences). We illustrate this approach by describing how it has been used to enable a robot to learn a difficult juggling task.”)
Gupta teaches:
train the machine learning algorithm to control at least one function of the controllable system, the machine learning algorithm being trained based on the first training data and the additional training data (Introduction, paragraph 1, “IoT devices have the potential to provide realtime, local, sensor-based solutions for a variety of areas like housing, factories, farming, even everyday utilities like toothbrushes and spoons. The ability to use machine learning on data collected from IoT sensors opens up a myriad of possibilities. For example, smart factories measure temperature, noise and various other parameters of their machines. ML based anomaly detection models [machine learning algorithms] can then be applied on this sensor data to preemptively schedule maintenance of a machine [controllable system] and avoid failure (i.e., wherein the ‘machine’ is interpreted as being the controllable system and the controlling function is interpreted as being ‘able to schedule maintenance of the machine’, hence that ‘ML based anomaly detection models can be applied to the data’ is interpreted as the controlling function is done based on the training)”)
A control device for generating training data for training a machine learning algorithm, (Introduction, paragraph 1, “Internet-of-things (IoT) is one such rapidly growing domain. IoT devices [control device] have the potential to provide realtime, local, sensor-based solutions for a variety of areas like housing, factories, farming, even everyday utilities like toothbrushes and spoons. The ability to use machine learning on data collected from IoT sensors (i.e., wherein it is interpreted that the device is generating data collected from the sensors on the control device) opens up a myriad of possibilities. For example, smart factories measure temperature, noise and various other parameters of their machines. ML based anomaly detection models [machine learning algorithm] can then be applied on this sensor data to preemptively schedule maintenance of a machine and avoid failure (i.e., wherein the data is then used for training)”)
A person of ordinary skill in the art would reasonably find the teachings of Gupta to be helpful in solving the problem of handling unlabeled data by using a control device to perform the repetitive tasks present in DeVries. In view of the teachings of Gupta it would have been obvious for a person of ordinary skill in the art to apply the teachings of Gupta to DeVries before the effective filing date of the claimed invention in order to improve the efficiency of computer resources (Gupta, Abstract, “ProtoNN models can be deployed even on devices with puny storage and computational power (e.g. an Arduino UNO with 2kB RAM) to get excellent prediction accuracy. ProtoNN derives its strength from three key ideas: a) learning a small number of prototypes to represent the entire training set, b) sparse low dimensional projection of data, c) joint discriminative learning of the projection and prototypes with explicit model size constraint. We conduct systematic empirical evaluation of ProtoNN on a variety of supervised learning tasks (binary, multi-class, multi-label classification) and show that it gives nearly state-of-the-art prediction accuracy on resource-scarce devices while consuming several orders lower storage, and using minimal working memory.”
Claim(s) 13-14 is/are rejected under 35 U.S.C. 103 as being unpatentable over DeVries et al., in view of Atkeson et al., further in view of Gupta et al., and Vogelsong et al., (US10926408B1)
Regarding claim 13 and analogous claim 14:
DeVries, as modified by Atkeson and Gupta, teaches the method of claim 1.
DeVries, as modified by Atkeson and Gupta, does not explicitly teach:
wherein the controllable system is one of (i) an injection system of an internal combustion engine, (ii) an analyzer configured to analyze samples for presence of viruses, (iii) a driver assistance system of a motor vehicle, (iv) a robotic system, (v) a kitchen appliance, or (vi) a washing machine
Vogelsong teaches:
wherein the controllable system is one of (i) an injection system of an internal combustion engine, (ii) an analyzer configured to analyze samples for presence of viruses, (iii) a driver assistance system of a motor vehicle, (iv) a robotic system, (v) a kitchen appliance, or (vi) a washing machine (Col 2-3, line 53, “By using targeted updates to machine-learned policies to control robotic task performance, the present technology is able to achieve levels of robustness, accuracy, and flexibility not available by traditional methods (i.e., wherein under the broadest reasonable interpretation (BRI) training the ml to control a controllable system (robotics system)). As an example, a machine learned robotic control policy may yield the capability to perform tasks that a human cannot figure out or imagine, for example an autopilot control policy that can recover from stall. Further, the disclosed targeted updates during policy training enable the disclosed machine learned robotic policies to achieve a greater level of robustness to scenarios outside of the training data, and can train a successful policy using fewer iterations and less time than traditional policy training techniques. Although aspects of some embodiments described in the disclosure will focus, for the purpose of illustration, on particular examples of training data, tasks, and robotic systems [a robotic system], the examples are illustrative only and are not intended to be limiting. In some embodiments, the techniques described herein may be applied to additional or alternative training data, robotic systems, and tasks. Various aspects of the disclosure will now be described with regard to certain examples and embodiments, which are intended to illustrate but not limit the disclosure.”)
A person of ordinary skill in the art would reasonably find the teachings of Vogelsong to be helpful in solving the problem of training a machine learning for using a robotic control system in DeVries. In view of the teachings of Vogelsong it would have been obvious for a person of ordinary skill in the art to apply the teachings of Vogelsong to DeVries before the effective filing date of the claimed invention in order to improve the efficiency of training data used in a controllable system (Vogelsong, Abstract, “Artificial intelligence describes computerized systems that can perform tasks typically considered to require human intelligence. The capability to learn is an important aspect of intelligence, as a system without this capability generally cannot become more intelligent from experience. Machine learning is a field of computer science that gives computers the ability to learn without being explicitly programmed”)
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, Any inquiry concerning this communication or earlier communications from the examiner should be directed to AMINA BENOURAIDA whose telephone number is (571)272-4340. The examiner can normally be reached Monday-Friday 8:30am-5pm 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, Michael J. Huntley can be reached at (303) 297-4307. 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.
/AMINA MORENO BENOURAIDA/ Examiner, Art Unit 2129
/MICHAEL J HUNTLEY/Supervisory Patent Examiner, Art Unit 2129
Prosecution Insights