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 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, 9, 11, 17 are rejected under 35 U.S.C. 103 as being unpatentable over Gao (US 20220101438 A1), in view of Chen (US 20200320769 A1), Agarwal (US 20240285255 A1), and Engelstein (US 20200333767 A1).
Regarding Claim 1, Gao discloses “receiving a plurality of inputs including domain parameters and initial weights; providing the plurality of inputs to an optimization model…”(See [0177]; input data is received and provided to a neural network for training)
“performing, using a first layer of the optimization model, a training and optimization process based on the plurality of inputs and based on a training objective;” (See [0172], [0174]; the neural network model is trained and optimized based on input data and objective)
“performing, using a second layer of the optimization model, a differencing operation on an output of the first layer;” (See [0182]; a differencing operation (loss function) is performed on the output of the previous performing step through backpropagation)
“calculating and storing, using a fourth layer of the optimization model, metrics regarding the training and optimization process;” (See [0183]; model metrics that are evaluated from the training process are stored. See also [0144]; the results of the process are requested to be stored to a repository)
Gao fails to explicitly disclose, “recording, using a third layer of the optimization model, a loss based on the training objective used by the optimization model”
“outputting, using the optimization model, updated weights.”.
Chen discloses “recording, using a third layer of the optimization model, a loss based on the training objective used by the optimization model” (See [0385], [0388]; the results of a loss function that is run on the model’s output are recorded).
“outputting, using the optimization model, updated weights.” (See [0238]; the weights are updated in the neural network, which are then outputted).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Gao by the implementation of recording loss based on the optimization model’s training objective that is taught by Chen, to make the invention record the results of the loss function operation; thus, one of ordinary skill in the art would be motivated to combine the references since it quantifies the "error" of a model by calculating a single numerical value representing the difference between predicted and actual target values, as Chen teaches that “For a parameter set Θ that is to be learnt, we then have the following error to minimize” (Chen, Paragraph [0385]).
Additionally, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Gao by the implementation of outputting updated weights from the optimization model that is taught by Chen, to make the invention update the weights in the model and then output the updated weights; thus, one of ordinary skill in the art would be motivated to combine the references since it presents users with updated information about the weights in the model, as Chen teaches that “This process will adjust the network weights of the pretrained neural network and repurpose the network to model the target garment image dataset” (Chen, Paragraph [0238]).
Gao-Chen fails to explicitly disclose, “wherein the first layer of the optimization model provides an output including a set of weights”.
Agarwal discloses “wherein the first layer of the optimization model provides an output including a set of weights” (See [0063]; Agarwal discloses that the first layer provides an output which includes weights).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Gao by the implementation of outputting a set of weights from the first layer that is taught by Agarwal, to make the invention output a set of weights for the second layer to take as an input, as Agarwal teaches “The first layer 702 provides an output f.sub.t, which includes weights that indicate which data from the previous cell state should be “forgotten” and which data from the previous cell state should be “remembered” by cell 700 (Agarwal, Paragraph [0063])”.
Gao-Chen-Agarwal fails to explicitly disclose, “wherein performing the differencing operation includes performing parameter-free transformations on the output of the first layer to obtain information related to each of a plurality of time periods”.
Engelstein discloses “wherein performing the differencing operation includes performing parameter-free transformations on the output of the first layer to obtain information related to each of a plurality of time periods” (See [0350]; Engelstein discloses performing a transformation on the output of the first layer by subtracting from the output, and pushes the resulting output to a time-series forecasting model algorithm to obtain information related to time periods. The transformation does not use any weights to perform subtraction).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Gao by the implementation of performing parameter-free transformations on the output of the first layer taught by Engelstein, to obtain information related to a plurality of time periods for each output, as Engelstein teaches “The gradient boosting takes the output of the first layer algorithm (i.e., the LASSO algorithm), subtracting from the expected output, and pushes the output to the second layer algorithm, e.g., a time-series forecasting model algorithm”.
Regarding claim 3, Gao discloses “using the first layer of the optimization model, the training and optimization process includes performing model fitting based on the plurality of inputs to fit a neural network with backpropagation” (See [0283]; a neural network can be trained using backpropagation with a loss function, as backpropagation is part of performing model fitting on a machine learning model).
Regarding claims 9 and 17, these claims are similar in scope to claim 1.
Regarding claim 11, this claim is similar in scope to claim 3.
Claim Rejections - 35 USC § 103
Claims 2, 10, 18 are rejected under 35 U.S.C. 103 as being unpatentable over Gao (US 20220101438 A1), in view of Chen (US 20200320769 A1), Agarwal (US 20240285255 A1), and Engelstein (US 20200333767 A1), and further in view of Lindgren (US 20220020077 A1).
Regarding claim 2, Gao fails to explicitly disclose “setting one or more hyperparameters of the optimization model, wherein the one or more hyperparameters include at least one of: whether the first layer of the optimization model is a dense layer or a recurrent layer; a number of hidden layers of the first layer of the optimization model; a number of units per hidden layer of the first layer of the optimization model; and a maximum number of epochs for the training objective”.
However, Lindgren discloses “setting one or more hyperparameters of the optimization model, wherein the one or more hyperparameters include at least one of: a number of hidden layers of the first layer of the optimization model; a number of units per hidden layer of the first layer of the optimization model; and a maximum number of epochs for the training objective” (See [0058]; this includes hyperparameters for the following: a number of hidden layers, a number of units per hidden layer, and a number of epochs).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Gao by specifying what hyperparameters to use for the optimization model that is taught by Lindgren, to control the invention’s process of training on given data; thus, one of ordinary skill in the art would be motivated to combine the references since it directly dictates how the optimization model learns and performs, as Lindgren teaches that “Hyperparameters that may be configured and adjusted between training sessions include, but are not limited to, the number of hidden layers and units, dropout (e.g., to avoid overfitting the model thus increasing the generalization power), network weight in initialization, the activation function (e.g., SoftMax, sigmoid, rectifier activation function, etc.), the learning rate (i.e., how quickly the network updates its parameters), momentum, number of epochs, batch size and the like” (Lindgren, Paragraph [0058]).
Regarding claim 10 and 18, this claim is similar in scope to claim 2.
Claim Rejections - 35 USC § 103
Claims 4, 12, 19 are rejected under 35 U.S.C. 103 as being unpatentable over Gao (US 20220101438 A1), in view of Chen (US 20200320769 A1), Agarwal (US 20240285255 A1), and Engelstein (US 20200333767 A1), and further in view of Ma (US 20200151288 A1).
Regarding claim 4, Gao fails to disclose that “performing the model fitting includes: adapting parameter learning rates in real time based on an average of first and second moments by calculating an exponential moving average of a gradient and a moving average of a squared gradient; controlling decay rates of the exponential moving average of the gradient and the moving average of the squared gradient; bias correcting one or more weight parameters; and updating the initial weights using the bias corrected one or more weight parameters”.
However, Ma discloses “performing the model fitting includes: adapting parameter learning rates in real time based on an average of first and second moments by calculating an exponential moving average of a gradient and a moving average of a squared gradient;” (See [0067]; the Adam algorithm is used to calculate the exponential moving average of a gradient and a combination of the AdaGrad and RMSProp algorithms are used to calculate the moving average of the squared gradient)
“controlling decay rates of the exponential moving average of the gradient and the moving average of the squared gradient;” (See [0067]; the Adam algorithm is also used to control the decay rates of the exponential moving average of the gradient and the moving average of the squared gradient)
“bias correcting one or more weight parameters;” (See [0067]; bias corrected estimates are derived from the decay rate parameters)
“and updating the initial weights using the bias corrected one or more weight parameters” (See [0067]; an optimization algorithm called an “Adam optimizer” is used to update weights in the network based on the data).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Gao by including the Adam, AdaGrad, and RMSProp equations; thus, one of ordinary skill in the art would be motivated to combine the references since they can be used to calculate the averages that are necessary for controlling the decay rates and updating the initial weights using bias corrected estimates that are derived from the decay rate parameters, as Ma teaches that an “‘Adam Optimizer’ refers to an optimization algorithm that can used instead of the classical stochastic gradient descent procedure to update network weights iterative based in training data. Stochastic gradient descent maintains a single learning rate (termed alpha) for all weight updates and the learning rate does not change during training. A learning rate is maintained for each network weight (parameter) and separately adapted as learning unfolds. Adam combines advantages of two other extensions of stochastic gradient descent: Adaptive Gradient Algorithm (AdaGrad) that maintains a per-parameter learning rate that improves performance on problems with sparse gradients (e.g. natural language and computer vision problems), and Root Mean Square Propagation (RMSProp) that also maintains per-parameter learning rates that are adapted based on the average of recent magnitudes of the gradients for the weight (e.g. how quickly it is changing). Adam also makes use of the average of the second moments of the gradients (the uncentered variance). Specifically, the algorithm calculates an exponential moving average of the gradient and the squared gradient, and the parameters beta l and beta 2 control the decay rates of these moving averages. The initial value of the moving averages and beta l and beta 2 values close to 1.0 (recommended) result in a bias of moment estimates towards zero. This bias is overcome by first calculating the biased estimates before then calculating bias-corrected estimates” (Ma, Paragraph [0067]).
Regarding claim 12 and 19, this claim is similar in scope to claim 4.
Claim Rejections - 35 USC § 103
Claims 5, 13 are rejected under 35 U.S.C. 103 as being unpatentable over Gao (US 20220101438 A1), in view of Chen (US 20200320769 A1), Agarwal (US 20240285255 A1), and Engelstein (US 20200333767 A1), and further in view of Marshall (US 20220122182 A1).
Regarding claim 5, Gao fails to explicitly disclose, “the plurality of inputs further includes final weight parameters, and wherein performing the training and optimization process includes performing a rebalancing process in which a path to achieve the final weight parameters is determined by the optimization model”.
However, Marshall discloses, “the plurality of inputs further includes final weight parameters, and wherein performing the training and optimization process includes performing a rebalancing process in which a path to achieve the final weight parameters is determined by the optimization model” (See [0076]; performing a rebalancing process can be done using a calculation engine, which is an algorithm that takes inputs to calculate estimates. As time passes, the calculation engine will automatically rebalance the spending plan (a path to achieve the final weight parameters) according to the given target weights (final weights). Also see [0059], the rebalancing model can be tested and trained by using machine learning).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Gao by the implementation of a rebalancing process that is taught by Marshall, to make the invention determine a path to achieve the final weight that was specified in the parameters; thus, one of ordinary skill in the arts would be motivated to combine the references since it shows how an optimization model can continuously make improvements to its portfolio to achieve the desired final weight parameters, as Marshall teaches that “The calculation engine 545 can also rebalance the individual spending plan as time passes and updated external data sets are available to the calculation engine 545. If the duration mismatch (calculated daily) between the individual spending plan and associated portfolio exceeds established thresholds the asset management tool 500 will run calculations detailed in step four above. The new portfolio will be compared to the existing portfolio and trades will be executed to rebalance to the new target weights” (Marshall, Paragraph [0076]) and also “Machine learning may be used to test and train the rebalance model used by the calculation engine” (Marshall, Paragraph [0059]).
Regarding claim 13, this claim is similar in scope to claim 5.
Claim Rejections - 35 USC § 103
Claims 6, 14 are rejected under 35 U.S.C. 103 as being unpatentable over Gao (US 20220101438 A1), in view of Chen (US 20200320769 A1), Agarwal (US 20240285255 A1), and Engelstein (US 20200333767 A1), and further in view of Renshaw (US 8533089 B1).
Regarding claim 6, Gao fails to explicitly disclose, “performing the training and optimization process includes performing a multi-period portfolio optimization, wherein the plurality of inputs includes returns data, risks data, volumes data, and the initial weights from a feeder model, and wherein the optimization model outputs multi-period weights for a defined time period”.
Chen discloses “the optimization model outputs multi-period weights” (See [0238]; weights in the neural network are updated, then outputted)
Therefore, it would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Gao by outputting updated weights from the optimization model that is taught by Chen, to make the invention update the weights in the model and then output the updated weights, as Chen teaches “For a parameter set Θ that is to be learnt, we then have the following error to minimize” (Chen, Paragraph [0385]).
Chen fails to explicitly disclose, “performing the training and optimization process includes performing a multi-period portfolio optimization, wherein the plurality of inputs includes returns data, risks data, volumes data, and the initial weights from a feeder model” and “for a defined time period”.
However, Renshaw discloses “performing the training and optimization process includes performing a multi-period portfolio optimization, wherein the plurality of inputs includes returns data, risks data, volumes data, and the initial weights from a feeder model” and “for a defined time period” (See col 9, line 57 through col 10, line 5; a financial model that uses a plurality of inputs that include historical information of returns data, risks data, and trade volumes data is shown).
Therefore, it would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to modify Chen by specifying a plurality of inputs and a defined time period that is taught by Renshaw. These inputs are parameters that are used for performing a multi-period portfolio optimization and specifying a defined time period limits the scope of the optimization, as Renshaw teaches “As illustrated in FIG. 3, and as described in greater detail below, additional inputs 30 may suitably include databases of historical data for back testing and the like, data sources for assets which may be included in portfolios, such as the asset symbols, tickers, or identification number, the current prices of stocks, bonds, commodities, currencies, options, other investment vehicles, and the like, data, such as current factors, risk models and return data, and the like. This data may also include historical information on macroeconomic variables, such as inflation and the rates for United States Treasury bonds of various maturities, for example. It will be recognized that a wide variety of additional inputs may be provided including without limitation other complementary or supplementary software, such as portfolio optimization modeling software, for example.” (Renshaw, col 9, line 57) and “As illustrated in FIG. 3, and as described in greater detail below, the system inputs 32 may suitably include the index universe, which defines a set of securities over which to define the factor index; a targeted factor, which defines a numerical value for each security in the universe; non-targeted factors, which define numerical values for other factors for the securities in the universe; risk models, which can be used to compute tracking errors; and data for the securities in the universe such as average daily trade volume, price, benchmark weight, and the like.” (Renshaw, col 10, line 5).
Regarding claim 14, this claim is similar in scope to claim 6.
Claim Rejections - 35 USC § 103
Claims 7, 15, 20 are rejected under 35 U.S.C. 103 as being unpatentable over Gao (US 20220101438 A1), in view of Chen (US 20200320769 A1), Agarwal (US 20240285255 A1), and Engelstein (US 20200333767 A1), and further in view of Renshaw (US 8533089 B1), and further in view of Serbin (US 20090157563 A1).
Regarding claim 7, Gao fails to explicitly disclose, “the multi-period portfolio optimization maximizes returns for a given forecasting based on market impact predictions”.
However, Serbin discloses “the multi-period portfolio optimization maximizes returns for a given forecasting based on market impact predictions” (See [0035], [0057]; a forecast of market impact is used to calculate expected returns for individual investment funds, and this can be used to determine how to maximize returns for that fund).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Gao by utilizing a market impact forecast to calculate expected returns as taught by Serbin, in order to determine the optimal amount of turnover for a fund to maximize returns for that fund, as Serbin teaches “The methods of the current invention can be executed with the perspective that a forecast of market-impact cost of a volume-weighted average price (VWAP) strategy is used in this case, rather than one tailored to more elaborate trading styles or venues, as it represents average liquidity demand by definition. As a building block for capacity analysis, daily market-impact cost estimates can be extrapolated, adjusted for leverage, cash drag, etc., to complete the net expected return calculation for individual funds.” (Serbin, Paragraph [0035]) and “Finally, as a goal of the invention is to determine the optimal amount of turnover for a given fund, i.e. the percentage of turnover for a given fund that results in the greatest net return on investment, the range of acceptable turnover must be defined” (Serbin, Paragraph [0057]).
Regarding claim 15 and 20, this claim is similar in scope to claim 7.
Claim Rejections - 35 USC § 103
Claims 8, 16 are rejected under 35 U.S.C. 103 as being unpatentable over Gao (US 20220101438 A1), in view of Chen (US 20200320769 A1), Agarwal (US 20240285255 A1), and Engelstein (US 20200333767 A1), and further in view of Renshaw (US 8533089 B1), and further in view of Choi (US 20220230244 A1).
Regarding claim 8, Gao fails to explicitly disclose, “performing at least one data consistency check on at least one input of the plurality of inputs, wherein the at least one data consistency check includes one or more of: checking that projected returns and volumes are in a certain range; checking that a variance-covariance matrix associated with the risks data is positive semidefinite; checking that the variance-covariance matrix associated with the risks data is positive definite; checking a consistency of a number of securities; checking validity of the volumes data; checking a feasibility of one or more constraints; and performing a boundless check on the plurality of inputs.”.
However, Choi discloses “performing at least one data consistency check on at least one input of the plurality of inputs, wherein the at least one data consistency check includes one or more of: checking validity of the volumes data.” (See [0105]; checking the validity of trading volumes is a mean of preventing forgery and falsification to ensure data consistency).
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Gao by performing a data consistency check on the inputs as taught by Choi, to ensure that forgery and falsification of the data is prevented, as Choi teaches “An energy trading method for an energy prosumer according to example embodiments may be a person-to-person trading system, and thus may resolve various security issues such as transparent disclosure of trading details, prevention of forgery and falsification of information such as a production volume and a trading volume, and prevention of trading denial through authentication between trading parties” (Choi, Paragraph [0105]).
Regarding claim 16, this claim is similar in scope to claim 8.
Response to Arguments
Applicant's arguments regarding the 35 USC 103 rejection are moot in view of the new grounds of rejection necessitated by applicant's amendments.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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 DAVID KIM whose telephone number is (571)272-4331. The examiner can normally be reached 7:30 AM - 4:30 PM.
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, Matthew Ell can be reached at (571) 270-3264. 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.
/D.K./Examiner, Art Unit 2141
/MATTHEW ELL/Supervisory Patent Examiner, Art Unit 2141
Prosecution Insights