DETAILED ACTION
This action is in response to the original filing on 12/14/2023. Claims 1-5 are pending and have been considered below.
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 .
Claim Objections
Claim 2 is objected to because of the following informalities: wherein the update necessity determination unit determines that the machine learning model needs to be updated when a duration of time during which the number of deviations is equal to or greater than an update determination threshold is equal to or greater than a continuation determination threshold. This limitation is unclear and appears as though it is missing separating punctuation or missing words. For the purposes of examination, the relevant terms of “update determination threshold” and “continuation determination threshold” are interpreted as separate and conditional statements, wherein the word “and” is placed after “update determination threshold” and before “is” in order to read “the number of deviations” as greater than or equal to both of these thresholds. Appropriate correction is required.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-5 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Claim 1
Step 1: Claim 1 recites a system and so is directed to the statutory category of a machine.
Step 2A Prong 1: The claim recites: calculates, based on the target vehicle data of the target vehicle in the area and the learning condition, the number of deviations that is the number of target vehicle data deviating from the learning condition; and under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of a mathematical calculation which is practically capable of being performed in the human mind with the assistance of pen and paper.
The claim further recites: determines the necessity of updating the machine learning model based on the number of deviations under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of an evaluation or prediction which is practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: The claim recites the additional elements of: a number-of-deviations calculation unit that and an update necessity determination unit that, which amount to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (MPEP 2106.05(h)). The claimed computer components are recited at a high level of generality and are merely invoked as a tool to perform the abstract idea.
Step 2B: The claim does not contain significantly more than the judicial exception. The claim recites the additional elements of: a number-of-deviations calculation unit that and an update necessity determination unit that, which amount to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (MPEP 2106.05(h)). The claimed computer components are recited at a high level of generality and are merely invoked as a tool to perform the abstract idea.
Claim 2
Step 1: Claim 2 recites a system, and so is directed to the statutory category of a machine.
Step 2A Prong 1: The claim recites: determines that the machine learning model needs to be updated when a duration of time during which the number of deviations is equal to or greater than an update determination threshold is equal to or greater than a continuation determination threshold under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of an evaluation or prediction which is practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: The claim recites the additional elements of: wherein the update necessity determination unit, which amount to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (MPEP 2106.05(h)). The claimed computer components are recited at a high level of generality and are merely invoked as a tool to perform the abstract idea.
Step 2B: The claim does not contain significantly more than the judicial exception. The claim recites the additional elements of: wherein the update necessity determination unit, which amount to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (MPEP 2106.05(h)). The claimed computer components are recited at a high level of generality and are merely invoked as a tool to perform the abstract idea.
Claim 3
Step 1: Claim 3 recites a system, and so is directed to the statutory category of a machine.
Step 2A Prong 1: The claim recites: calculates, as the number of deviations, the number of unstable deviations that is the number of target vehicle data deviating from the learning condition when the unstable behavior of the target vehicle occurs under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of a mathematical calculation which is practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: The claim recites the additional elements of: the machine learning model predicts an occurrence of unstable behavior that is a sudden change in behavior of the target vehicle; and the number-of-deviations calculation unit which amount to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (MPEP 2106.05(h)). The claimed computer components are recited at a high level of generality and are merely invoked as a tool to perform the abstract idea.
Step 2B: The claim does not contain significantly more than the judicial exception. The claim recites the additional elements of: the machine learning model predicts an occurrence of unstable behavior that is a sudden change in behavior of the target vehicle; and the number-of-deviations calculation unit which amount to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (MPEP 2106.05(h)). The claimed computer components are recited at a high level of generality and are merely invoked as a tool to perform the abstract idea.
Claim 4
Step 1: Claim 4 recites a system, and so is directed to the statutory category of a machine.
Step 2A Prong 1: The claim recites: predicts an occurrence of unstable behavior that is a sudden change in behavior of the target vehicle; under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of an evaluation or prediction which is practically capable of being performed in the human mind with the assistance of pen and paper.
The claim further recites: calculates, as the number of deviations, the number of unstable deviations that is the number of target vehicle data deviating from the learning condition when the unstable behavior of the target vehicle occurs and the number of normal deviations that is the number of target vehicle data deviating from the learning condition when the unstable behavior does not occur; under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of a mathematical calculation which is practically capable of being performed in the human mind with the assistance of pen and paper.
The claim further recites: the update necessity determination unit determines that the machine learning model needs to be updated when a duration of time during which the number of unstable deviations is equal to or greater than an update determination threshold is equal to or greater than a continuation determination threshold; and under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of an evaluation or prediction which is practically capable of being performed in the human mind with the assistance of pen and paper.
The claim further recites: when the number of normal deviations is equal to or greater than an early determination threshold, the update necessity determination unit sets the continuation determination threshold to a smaller value than the continuation determination threshold when the number of normal deviations is less than the early determination threshold. Under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of a mathematical calculation which is practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: The claim recites the additional element of: the machine learning model, and the number-of-deviations calculation unit which amount to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (MPEP 2106.05(h)). The claimed computer components are recited at a high level of generality and are merely invoked as a tool to perform the abstract idea.
Step 2B: The claim does not contain significantly more than the abstract idea. The claim recites the additional element of: the machine learning model, and the number-of-deviations calculation unit which amount to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (MPEP 2106.05(h)). The claimed computer components are recited at a high level of generality and are merely invoked as a tool to perform the abstract idea.
Claim 5
Step 1: Claim 5 recites a method, and so is directed to the statutory category of a method.
Step 2A Prong 1: The claim recites: calculating, based on the target vehicle data of the target vehicle in the area, the number of deviations that is the number of target vehicle data deviating from the learning condition; and under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of a mathematical calculation which is practically capable of being performed in the human mind with the assistance of pen and paper.
The claim further recites: determining the necessity of updating the machine learning model based on the number of deviations under its broadest reasonable interpretation in light of the specification, this limitation encompasses the mental process of determination and judgement which is practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2 & Step 2B: There are no additional elements recited, so the claim does not provide a practical application and is not considered to be significantly more. As such, the claims are patent ineligible.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1-2 and 5 are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Delchev et al. (US 20240185172 A1, hereinafter Delchev).
Regarding Claim 1, Delchev teaches a model update necessity determination system that is a system that determines necessity of updating a machine learning model, the machine learning model being a model with which learning is performed using vehicle data under a predetermined learning condition and that predicts, based on target vehicle data acquired from a target vehicle within a preset area, a behavior change of the target vehicle, the model update necessity determination system comprising: (Paragraph [0005] The process described herein uses at least two different machine learning models: a first trained machine learning model configured to output a price prediction for automobile transportation at the present time, and a second trained machine learning model configured to output a prediction of how an automobile transportation price will change over a time period. When a user provides a request for automobile transportation via a user interface, additional data (e.g., information about the vehicle to be moved) is collected, and then data from that request and additional data is provided to the first trained machine learning model. Paragraph [0006] The computing device may receive, as output from the second trained machine learning model, a predicted price change. The predicted price change may comprise a comparison between the transportation time and a time when the request for automobile transportation is received. Paragraph [0072] As part of step 501 and/or step 502, the one or more machine learning servers 302a may perform data collection for the purposes of generating training data. Data collection may comprise automated process that collects data for already completed vehicle transportation activities. Such collected data might be limited to a timeframe, such as from 2013 to the present time, as such a time range ensures that the most recent orders are included in the data. Such data may include, GPS coordinate information (e.g., GPS coordinates of the origin location and destination location) Paragraph [0073] Such processing may be performed because the market behavior is not the same across all U.S. states, because the number of orders is large for some U.S. states but there are some states with few data points, and/or because price predictions might not be accurate for orders coming from states with relatively fewer orders. (Behavior change of the target vehicle, the change being a change in price. The update occurring under various pre-requisite, one such being when crossing U.S. state lines [0073]. Targeted vehicle data is retrieved from varying sources, once such being actual GPS coordinates corresponding to a vehicle.)), including:
a number-of-deviations calculation unit that calculates, based on the target vehicle data of the target vehicle in the area and the learning condition, the number of deviations that is the number of target vehicle data deviating from the learning condition; and (Paragraph [0007] The computing device may then, based on determining that the predicted automobile transportation price prediction trend satisfies a threshold (e.g., suggesting that the trained machine learning model is skewing in an undesirable direction, such as predicting prices too high), discard (e.g., stop using) the first trained machine learning model, update the first training data by modifying one or more portions of the first training data (e.g., removing high values in the first training data), and generate a new version of the first trained machine learning model by training, based on the updated first training data, the first machine learning model to output the price prediction for automobile transportation. Paragraph [0075] To prevent such disbalance, if a fraction is above a specific threshold, the training data may be modified to, for example, select greater or fewer data points with higher or lower prices. Once modified, the final prices may be adjusted automatically upwards or downwards keeping the overall model error within reasonable limits. (The device determines how deviant/skewed training data has become by determining a fraction of training data above a threshold in order to update the machine learning model.))
an update necessity determination unit that determines the necessity of updating the machine learning model based on the number of deviations. (Paragraph [0007] The computing device may then, based on determining that the predicted automobile transportation price prediction trend satisfies a threshold... generate a new version of the first trained machine learning model by training, based on the updated first training data, the first machine learning model to output the price prediction for automobile transportation. Paragraph [0008] The computing device may periodically update one or more of the trained machine learning models. Paragraph [0075] Leaning too much in either direction may therefore cause problems and discourage reliance on the machine learning model(s). To prevent such disbalance, if a fraction is above a specific threshold, the training data may be modified to, for example, select greater or fewer data points with higher or lower prices. (Updating the model based on an amount of deviations surpassing a specified threshold)).
Regarding Claim 2, Delchev teaches: wherein the update necessity determination unit determines that the machine learning model needs to be updated when a duration of time during which the number of deviations is equal to or greater than an update determination threshold is equal to or greater than a continuation determination threshold. (Paragraph [0007] That new version of the first trained machine learning model may be thereby trained to avoid the undesirable predicted automobile transportation price prediction trend. Paragraph [0008] For example, the computing device may determine a time associated with the generation of the first trained machine learning model. Then, based on determining that the time satisfies a threshold, the computing device may discard (e.g., stop using) the first trained machine learning model, update the first training data (e.g., to contain newer data), and generate a new version of the first trained machine learning model by training, based on the updated first training data, the first machine learning model to output the price prediction for automobile transportation. Paragraph [0075] To prevent such disbalance, if a fraction is above a specific threshold, the training data may be modified to, for example, select greater or fewer data points with higher or lower prices. Once modified, the final prices may be adjusted automatically upwards or downwards keeping the overall model error within reasonable limits. (Update determination based on a set amount of time passed. The model is trained to avoid undesirable trends by periodic retraining, particularly if conditions surpass a determined threshold in order to reduce errors in the model.))
Regarding Claim 5, the claim is a method claim that corresponds to the system of Claim 1, and so is rejected for the same reasons.
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 3 and 4 are rejected under 35 U.S.C. 103 as being unpatentable over Delchev, as applied in the rejection of Claim 1 above, in view of Gama et al (‘Learning with Drift Detection’, September 29, 2004, hereinafter Gama).
Regarding Claim 3, Delchev teaches all of the limitations of Claim 1. However, Delchev fails to teach: the machine learning model predicts an occurrence of unstable behavior that is a sudden change in behavior of the target vehicle; and the number-of-deviations calculation unit calculates, as the number of deviations, the number of unstable deviations that is the number of target vehicle data deviating from the learning condition when the unstable behavior of the target vehicle occurs.
In the same field of endeavor, Gama teaches: the machine learning model predicts an occurrence of unstable behavior that is a sudden change in behavior of the target vehicle; and (Abstract When a new training example is available, it is classified using the actual model. Statistical theory guarantees that while the distribution is stationary, the error will decrease. When the distribution changes, the error will increase. Section 3 The method controls the trace of the online error of the algorithm. Suppose a sequence of examples where the error of the actual model increases reaching the warning level at example kw, and the drift level at example kd. This is an indication of a change in the distribution of the examples. A new context is declared starting in example kw, and a new decision model is induced using only the examples starting in kw till kd. (Declaring a new context upon an increase in the monitored error of the machine learning model, which is an occurrence of unstable behavior that is a sudden change in behavior)) and
the number-of-deviations calculation unit calculates, as the number of deviations, the number of unstable deviations that is the number of target vehicle data deviating from the learning condition when the unstable behavior of the target vehicle occurs. (Abstract The idea behind the drift detection method is to control the online error-rate of the algorithm. The training examples are presented in sequence. When a new training example is available, it is classified using the actual model. Statistical theory guarantees that while the distribution is stationary, the
error will decrease. When the distribution changes, the error will increase. The method controls the trace of the online error of the algorithm. For the actual context we define a warning level, and a drift level. A new context is declared, if in a sequence of examples, the error increases reaching the warning level at example kw, and the drift level at example kd. (The count of erroneous examples in the system (the number of unstable deviations) accrues while the error increases (when the unstable behavior occurs).)
It would have been obvious to one having ordinary skill in the art to have combined the references of Delchev and Gama before the effective filing date of the claimed invention because the two are analogous art and within the same field of endeavor, in specific maintaining and updating a deployed machine learning prediction model. These two address the same problem of determining when a model's predictions have degraded enough to require retraining and as such a combination of Gama's error rate-based drift detection with Delchev's price prediction model and its contemplated drift-monitoring module (Delchev [0074]), would have been obvious before the effective filing date. This modification would have been motivated by the desire to obtain statistically grounded, timely detection of model degradation, as Delchev itself calls for "a custom monitoring module … for detecting drifts" (Delchev [0074]; Gama Abstract, Section 3).
Regarding Claim 4, Delchev teaches all of the limitations of Claim 1. However, Delchev fails to teach: the machine learning model predicts an occurrence of unstable behavior that is a sudden change in behavior of the target vehicle; the number-of-deviations calculation unit calculates, as the number of deviations, the number of unstable deviations that is the number of target vehicle data deviating from the learning condition when the unstable behavior of the target vehicle occurs and the number of normal deviations that is the number of target vehicle data deviating from the learning condition when the unstable behavior does not occur; the update necessity determination unit determines that the machine learning model needs to be updated when a duration of time during which the number of unstable deviations is equal to or greater than an update determination threshold is equal to or greater than a continuation determination threshold; and when the number of normal deviations is equal to or greater than an early determination threshold, the update necessity determination unit sets the continuation determination threshold to a smaller value than the continuation determination threshold when the number of normal deviations is less than the early determination threshold.
In the same field of endeavor, Gamu teaches: the machine learning model predicts an occurrence of unstable behavior that is a sudden change in behavior of the target vehicle; (Section 3 We use a warning level to define the optimal size of the context window. The context window will contain the old examples that are on the new context and a minimal number of examples on the old context. Suppose that in the sequence of examples that traverse a node, there is an example i with correspondent pi and si… Suppose a sequence of examples where the error of the actual model increases reaching the warning level at example kw, and the drift level at example kd. This is an indication of a change in the distribution of the examples. A new context is declared starting in example kw, and a new decision model is induced using only the examples starting in kw till kd. (The system describes an increase in the models monitored error signals that the distribution “changes”. After, it declares a new context. This comprises an occurrence of unstable behavior.))
the number-of-deviations calculation unit calculates, as the number of deviations, the number of unstable deviations that is the number of target vehicle data deviating from the learning condition when the unstable behavior of the target vehicle occurs and the number of normal deviations that is the number of target vehicle data deviating from the learning condition when the unstable behavior does not occur; (Section 3 For sufficient large values of the example size, the Binomial distribution is closely approximated by a Normal distribution with the same mean and variance. Considering that the probability distribution is unchanged when the context is static, then the 1 – α/2 confidence interval for p with n > 30 examples is approximately pi +- α * si. The parameter α depends on the confidence level. The drift detection method manages two registers during the training of the learning algorithm, pmin and smin. Every time a new example i is processed those values are updated when pi + si is lower than pmin + smin. (The registers of pmin and smin track the deviation rate updated while the distribution rate remains stationary (the number of normal deviations), versus the elevated error counted as the warning and drift levels are approached (the number of unstable deviations))
the update necessity determination unit determines that the machine learning model needs to be updated when a duration of time during which the number of unstable deviations is equal to or greater than an update determination threshold is equal to or greater than a continuation determination threshold; and (Section 3 Suppose that in the sequence of examples that traverse a node, there is an example i with correspondent pi and si. In the experiments described below the confidence level for warning has been set to 95%, that is, the warning level is reached if pi + si >= pmin +2 * smin. The confidence level for drift has been set to 99%, that is, the drift level is reached if pi + si >= pmin + 3 * smin. Suppose a sequence of examples where the error of the actual model increases reaching the warning level at example kw, and the drift level at example kd. This is an indication of a change in the distribution of the examples. A new context is declared starting in example kw, and a new decision model is induced using only the examples starting in kw till kd. It is possible to observe an increase of the error reaching the warning level, followed by a decrease. We assume that such situations correspond to a false alarm, without changing the context. Figure 1 details the dynamic window structure. With this method of learning and forgetting we ensure a way to continuously keep a model better adapted to the present context. (The drift of data is declared, and a new model is learned, when, over a sequence of examples, the error reaches the warning level at example kw and persists to the drift level at example kd (a duration during which the deviation count stays at or above a threshold reaches a further threshold)))
when the number of normal deviations is equal to or greater than an early determination threshold, the update necessity determination unit sets the continuation determination threshold to a smaller value than the continuation determination threshold when the number of normal deviations is less than the early determination threshold. (Section 3 We use a warning level to define the optimal size of the context window. The context window will contain the old examples that are on the new context and a minimal number of examples on the old context. Suppose that in the sequence of examples that traverse a node, there is an example i with correspondent pi and si. In the experiments described below the confidence level for warning has been set to 95%, that is, the warning level is reached if pi + si >= pmin +2 * smin. The confidence level for drift has been set to 99%, that is, the drift level is reached if pi + si >= pmin + 3 * smin. Suppose a sequence of examples where the error of the actual model increases reaching the warning level at example kw, and the drift level at example kd. This is an indication of a change in the distribution of the examples. A new context is declared starting in example kw, and a new decision model is induced using only the examples starting in kw till kd. It is possible to observe an increase of the error reaching the warning level, followed by a decrease. We assume that such situations correspond to a false alarm, without changing the context. Figure 1 details the dynamic window structure. With this method of learning and forgetting we ensure a way to continuously keep a model better adapted to the present context. (As stated in the limitation prior, the warning level "pi + si ≥ pmin + 2 ∗ smin" is reached before, and is a lower trigger than, the drift level "pi + si ≥ pmin + 3 ∗ smin," so crossing the early (warning) level reduces the further persistence needed to confirm drift, analogous to setting the continuation determination threshold to a smaller value)).
It would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to have incorporated the concepts of predicting occurrences of unstable behavior, calculating deviations, determining update necessity, and determining that if the deviations surpass a given threshold, then updating the model accordingly as taught by Gama into Delchev as both are analogous art and within the same field of endeavor, specifically maintaining and updating deployed machine learning prediction models. It would have been obvious before the effective filing date to have incorporated Gama’s two level warning and drift detection with Delchev’s price prediction model and its drift monitoring module as doing so would address the need in the art to detect these variances and would allow for the usage of stable-period errors to serve as an early indicator that shortens time between necessary model retraining periods, thereby limiting model degradation while avoiding re-training before necessary (Delchev [0074], Gama Section 3).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicants’ disclosure.
Takhirov et al. (US 11866056 B2) discusses anomaly tracking in vehicles.
WHYLIE et al. (US-20210056778-A1) discusses tracking of physical anomalies in vehicle tires.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JUSTIN A CARDOSO whose telephone number is (571)272-8512. The examiner can normally be reached M-F 7:30 - 5:00, alternate Friday's off.
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, Jennifer Welch can be reached at (571) 272-7212. 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.
/JUSTIN CARDOSO/
Patent Examiner, Art Unit 2143
/JENNIFER N WELCH/Supervisory Patent Examiner, Art Unit 2143
Prosecution Insights