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 .
This action is made non-final.
Claims 1-20 are pending. Claims 1, 3 and 12 are independent claims.
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 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding claim 1:
Step 1: This part of the eligibility analysis evaluates whether the claim falls within any statutory category. See MPEP 2106.03. Claim 1 recites: A system, comprising: a processor in communication with a memory, the memory including instructions executable by the processor to: Claim 1 is directed to an apparatus (Step 1: YES).
Step 2A prong 1: Does the claim recite a judicial exception? Claim 1 recites: apply an example microstate instance of a microstate space as input… to obtain an example macrostate of the example microstate instance (under the broadest reasonable interpretation (BRI) of microstate and macrostate, microstates can be lower level data descriptors, and macrostates can be higher level data descriptors – a human is capable of mapping data to higher level representations, for example, classification of an image or text, i.e., relating a cat image to a specific breed/coloring pattern, or facial recognition, i.e., relating facial features to a known person, create a mapping from data descriptors to higher level data descriptors), and sample… an ensemble of sampled microstate instances of the first microstate space or the second microstate space that correspond to the example macrostate (sampling, i.e., selecting, an ensemble of microstate instances, is interpreted under BRI as selecting data, which is a mental process – i.e., selecting data that correspond to a criteria). These steps can be performed mentally or are mathematical calculations (Step 2A prong 1: YES).
Step 2A prong 2: Does the claim recite additional elements? Do those additional elements, considered individually and in combination, integrate the judicial exception into a practical application? Claim 1 recites: to a neural network… the neural network being one of a first neural network having learned a first mapping between a first microstate space of a microstate pair and a first macrostate, and a second neural network having learned a second mapping between a second microstate space of the microstate pair and a second macrostate; the first microstate space and the second microstate space each including observation data about a physical system, where the first microstate space corresponds with a first type of observation data about the physical system, and where the second microstate space corresponds with a second type of observation data about the physical system… by the neural network… the neural network being an invertible neural network. Specifying that the neural network is trained to perform a task without a) placing limits on how the model operates or b) providing details on how the model is trained, is an attempt to use the neural network model by merely applying the abstract idea (i.e. perform the mental process on a generic computer). Further, the claim omits any details as to how the neural network solves a technical problem and instead only recites the idea of a solution or outcome. Thus, the limitation represents no more than mere instructions to implement the abstract idea which is equivalent to adding the words “apply it” to the recited judicial exception. See MPEP 2106.05(f). Specifying that the neural network is used to perform the mental processes of outputting a macrostate based on an input microstate and sampling an ensemble of sampled microstate instances, is also mere instructions to implement the abstract idea on a generic computer. Specifying that the neural network is invertible, without describing any structural components or reference to how the invertible nature of the model solves a technical problem, is also mere instructions to apply the judicial exception. Specifying that the microstate spaces include different types of physical system observation data is an additional element specifying a field of use without significantly more (Step 2A prong 2: NO).
Step 2B: These elements are recited at such a high level of generality that they fail to integrate the abstract idea into a practical application, since they provide nothing more than mere instructions to implement an abstract idea on a generic computer (MPEP 2106.05(f)), or limit the field of use without significantly more (MPEP 2106.05(h)). These limitations, taken either alone or in combination, fail to provide an inventive concept (Step 2B: NO). Thus, the claim is not patent eligible.
Regarding claim 2, it recites limitations which further narrow the abstract idea by specifying more details of the mental and mathematical process that occurs (Claim 2, providing a set of paired microstate space training data is insignificant extra-solution activity of data gathering or selecting a particular data source or type of data to be manipulated (see MPEP 2106.05(g)), and jointly training machine learning models in a way that a difference between mappings is minimized is still an attempt to apply the judicial exception, since it omits details as to how the neural network solves a technical problem. Additionally, there is also no constraint that the microstate pairs are related to the same macrostate. A mapping difference being minimized is similar to how a human can mentally map paired image and caption data to the same movie scene or mentally map word/meanings in two foreign languages back and forth with their native language).
Regarding claim 3:
Step 1: This part of the eligibility analysis evaluates whether the claim falls within any statutory category. See MPEP 2106.03. Claim 3 recites: A method, comprising… Claim 3 is directed to a method (Step 1: YES).
Step 2A prong 1: Does the claim recite a judicial exception? Claim 3 recites: applying an example microstate instance of a microstate space as input… to obtain an example macrostate of the example microstate instance, (a human is capable of mentally mapping data to higher level representations, for example, mental classification of an image or text, i.e., relating a cat image to a specific breed/coloring pattern, or mental facial recognition, i.e., relating facial features to a known person) and sampling… an ensemble of sampled microstate instances of the first microstate space or the second microstate space that correspond to the example macrostate (sampling, i.e., selecting, an ensemble of microstate instances, is interpreted under BRI as selecting data, which is a mental process). These steps can be performed mentally or are mathematical calculations (Step 2A prong 1: YES).
Step 2A prong 2: Does the claim recite additional elements? Do those additional elements, considered individually and in combination, integrate the judicial exception into a practical application? Claim 3 recites: to an neural network… the neural network being one of a first neural network having learned a first mapping between a first microstate space of a microstate pair and a first macrostate, and a second neural network having learned a second mapping between a second microstate space of the microstate pair and a second macrostate… by the neural network… the neural network being an invertible neural network. Specifying that the neural network is trained to perform a task without a) placing limits on how the model operates or b) providing details on how the model is trained, is an attempt to use the neural network model by merely applying the abstract idea (i.e. perform the mental process on a generic computer). Further, the claim omits any details as to how the neural network solves a technical problem and instead only recites the idea of a solution or outcome. Thus, the limitation represents no more than mere instructions to implement the abstract idea which is equivalent to adding the words “apply it” to the recited judicial exception. See MPEP 2106.05(f). Specifying that the neural network is used to perform the mental processes of outputting a macrostate based on an input microstate and sampling an ensemble of sampled microstate instances, is also mere instructions to implement the abstract idea on a generic computer. Specifying that the neural network is invertible, without describing any structural components or reference to how the invertible nature of the model solves a technical problem, is also mere instructions to apply the judicial exception (Step 2A prong 2: NO).
Step 2B: These elements are recited at such a high level of generality that they fail to integrate the abstract idea into a practical application, since they provide nothing more than mere instructions to implement an abstract idea on a generic computer (MPEP 2106.05(f)). These limitations, taken either alone or in combination, fail to provide an inventive concept (Step 2B: NO). Thus, the claim is not patent eligible.
Regarding claims 4-11, they recite limitations which further narrow the abstract idea by specifying more details of the mental and mathematical process that occurs (Claim 4, sampling outputs by inputting data into an invertible network is still mere instructions to apply the abstract idea without significantly more (MPEP 2106.05(f)), omitting details on how the sampling process or invertible nature of the network solves a technical problem and model structural constraints; Claim 5, inverting an invertible neural network is also an attempt to apply the abstract idea without significantly more, and omits details on how the invertible nature of the network solves a technical problem; Claim 6, similar to claim 4, sampling outputs by inputting data into an invertible network is still mere instructions to apply the abstract idea without significantly more (MPEP 2106.05(f)), omitting details on how the sampling process or invertible nature of the network solves a technical problem and model structural constraints; Claim 7, inverting an invertible neural network is also an attempt to apply the abstract idea without significantly more, and omits details on how the invertible nature of the network solves a technical problem; Claim 8, specifying that the two microstate spaces include different types of observation data about a physical system is an attempt to limit the field of use of the invention without significantly more (MPEP 2106.05(h)); Claim 9, specifying that the microstate space pair follows a joint distribution that relates the first microstate space to the second is insignificant extra-solution activity of selecting particular types of data to be manipulated – see MPEP 2106.05(g); Claim 10, providing a set of paired microstate space training data is insignificant extra-solution activity of data gathering or selecting a particular data source or type of data to be manipulated (see MPEP 2106.05(g)), and jointly training machine learning models in a way that a difference between mappings is minimized is still an attempt to apply the judicial exception, since it omits details as to how the neural network solves a technical problem and recites only the idea of a solution or outcome i.e., the claim fails to recite details of how a solution to a problem is accomplished; Claim 11, sampling, i.e., selecting, microstates based on some criteria, i.e., an input macrostate, is a mental process. Data selected based on some criteria follows a distribution that is contingent on those criteria. A human can mentally determine/select/brainstorm data that differs depending on provided criteria).
Regarding claim 12:
Step 1: This part of the eligibility analysis evaluates whether the claim falls within any statutory category. See MPEP 2106.03. Claim 12 recites: A method, comprising… Claim 12 is directed to a method (Step 1: YES).
Step 2A prong 1: Does the claim recite a judicial exception? Claim 12 recites: learn a first mapping between the first microstate space and a first macrostate and… learn a second mapping between the second microstate space and a second macrostate using the set of training data (a human is capable of mentally mapping data to higher level representations, for example, mental classification of an image or text, i.e., relating a cat image to a specific breed/coloring pattern, or mental facial recognition, i.e., relating facial features to a known person). These steps can be performed mentally or are mathematical calculations (Step 2A prong 1: YES).
Step 2A prong 2: Does the claim recite additional elements? Do those additional elements, considered individually and in combination, integrate the judicial exception into a practical application? Claim 12 recites: providing a set of training data as input to a first neural network and a second neural network, the set of training data including a plurality of training microstate pairs, where each respective training microstate pair of the plurality of training microstate pairs includes a first training microstate instance belonging to a first microstate space and a second training microstate instance belonging to a second microstate space; the first microstate space and the second microstate space each including observation data about a physical system, where the first microstate space corresponds with a first type of observation data about the physical system, and where the second microstate space corresponds with a second type of observation data about the physical system; and jointly training the first neural network to… training the second neural network to… such that a difference between the first macrostate and the second macrostate is minimized for each training microstate pair of the plurality of training microstate pairs of the set of training data. Providing a set of training data consisting of pairs of microstate instances from a first and second microstate space is insignificant extra-solution activity of data gathering that does not add a meaningful limitation to the training process. Specifying that the microstate spaces each include observation data about a physical system, of a first type and second type, is an additional element specifying a field of use without significantly more. Jointly training the two neural networks so that a difference between macrostate mappings of the microstate instances is minimized is reciting only the idea of a solution or outcome i.e., the claim fails to recite details of how a solution to a problem is accomplished. Thus, the limitation represents no more than mere instructions to implement the abstract idea which is equivalent to adding the words “apply it” to the recited judicial exception. See MPEP 2106.05(f). Specifying that the neural network is used to perform the mental processes of outputting a macrostate based on an input microstate and sampling an ensemble of sampled microstate instances, is also mere instructions to implement the abstract idea on a generic computer. Specifying that the neural network is invertible, without describing any structural components or reference to how the invertible nature of the model solves a technical problem, is also mere instructions to apply the judicial exception (Step 2A prong 2: NO).
Step 2B: These elements are recited at such a high level of generality that they fail to integrate the abstract idea into a practical application, since they only amount to data gathering or outputting without significantly more (MPEP 2106.05(g)), limit the field of use without significantly more, (MPEP 2106.05(h)), or provide nothing more than mere instructions to implement an abstract idea on a generic computer (MPEP 2106.05(f)). These limitations, taken either alone or in combination, fail to provide an inventive concept (Step 2B: NO). Thus, the claim is not patent eligible.
Regarding claims 13-16, they recite limitations which further narrow the abstract idea by specifying more details of the mental and mathematical process that occurs (Claim 13, describing an iterative loss function minimization process is a series of mathematical calculations, and specifying that the microstate spaces of training pairs are related to each other via a joint distribution is insignificant extra-solution activity of selecting a particular data source or type of data to be manipulated (MPEP 2106.05(g)); Claim 14, describing an iterative loss function minimization process is a series of mathematical calculations, and incorporating a distribution loss in the loss function that enforces mappings to have non-zero Jacobian determinants recites mathematical calculations; Claim 15, sampling (i.e., selecting) microstates based on some criteria (i.e., an input macrostate) is a mental process. Data selected based on some criteria follows a distribution that is contingent on those criteria. A human can mentally determine/select/brainstorm data that differs depending on provided criteria; Claim 16, using neural networks to perform a mentally performable microstate-macrostate mapping process (discussed in relation to claim 1 – see above), and performing a mentally performable sampling process (also discussed in relation to claim 1 – see above) is an attempt to apply the judicial exception without significantly more).
Regarding claims 17-20, they recite similar limitations to claims 4-7, respectively, and are rejected on the same grounds – see above.
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.
Claim(s) 1-8, 10-12, and 15-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Zhang et al. (“Probabilistic invertible neural network for inverse design space exploration and reasoning”, 2022), herein Zhang, in view of Keysers et al. (US 20240153256 A1), herein Keysers.
Regarding claim 1, Zhang teaches: A system, comprising:… apply an example microstate instance of a microstate space as input to a neural network to obtain an example macrostate of the example microstate instance (pg. 4, fig. 1, microstate x and macrostate y), the neural network being one of a first neural network having learned a first mapping between a first microstate space… and a first macrostate (pg. 4, see figure 1 and caption, Obtain y and uncertainty from x in forward process. The posterior distribution of x from a specific y is acquired by sampling z and fixing σ using the inverse process – and – pg. 9, We train the P-INN – P-INN stands for probabilistic invertible neural network)… the first microstate space… each including observation data about a physical system, where the first microstate space corresponds with a first type of observation data about the physical system (pg. 15, It is natural for us to apply P-INN to the inverse design of the blade – and – pg. 16, The simulated 3D structure of the turbine blade is shown in Figure 6. In reality, the stresses and strains of the blade are affected by a combination of pressure and thermal effects. We first calculate the stress and strain that are only affected by the pressure while ignoring the thermal effect; then analyze the stress and strain that only has the thermal effect while ignoring the pressure; finally, the simulation model with the combined effect of the two factors is obtained. The stress or temperature heat maps of the three processes are shown in Figure 7. The pressure model parameters include Young's modulus, Poisson’s ratio nu, the thermal expansion coefficient of the nickel-chromium alloy, and the pressures 1 and 2 on both sides of the blade; the thermal effect model parameters include the thermal conductivity of the nickel-chromium alloy – also see the referenced figures 6 and 7)… and sample, by the neural network, an ensemble of sampled microstate instances of the first microstate space… that correspond to the example macrostate, the neural network being an invertible neural network (pg. 6, The inverse process takes response y, predefined σ, and the sampled z as input, and derives the x… A one-to-many mapping of the inverse process (a set of y generates multiple sets of x) is implemented by sampling z multiple times).
Zhang fails to explicitly teach: a processor in communication with a memory, the memory including instructions executable by the processor to… of a microstate pair… and a second neural network having learned a second mapping between a second microstate space of the microstate pair and a second macrostate… and the second microstate space… and where the second microstate space corresponds with a second type of observation data about the physical system… or the second microstate space.
However, in the same field of endeavor, Keysers teaches: a processor in communication with a memory, the memory including instructions executable by the processor to (¶4, a system may include a processor and a non-transitory computer-readable medium having stored thereon instructions that, when executed by the processor, cause the processor to perform operations – Zhang never explicitly discloses any hardware)… of a microstate pair… and a second neural network having learned a second mapping between a second microstate space of the microstate pair and a second macrostate (¶2, A first machine learning model and a second machine learning model, each configured to process a different type of data, may be trained using a contrastive learning process to generate similar latent representation for matched pairs of input samples of the different types of data)… and the second microstate space… and where the second microstate space corresponds with a second type of observation data about the physical system (¶2, A first machine learning model and a second machine learning model, each configured to process a different type of data, may be trained using a contrastive learning process to generate similar latent representation for matched pairs of input samples of the different types of data)… or the second microstate space (Zhang discusses a single model approach, with a loss function comprising multiple parts. A person of ordinary skill in the art could train two of Zhang’s P-INN models on different input data types, similar to Keyser, and add an additional contrastive loss term to the loss functions of each model).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to train another similar network that operates on a different type of data (using a system comprising a processor and memory) as disclosed by Keysers in the system disclosed by Zhang to enable the trained models to easily adapt to new tasks (¶24, ML models trained in this manner may be configured to perform new tasks without additional data (i.e., zero-shot learning, sometimes referred to as zero-shot transfer) and/or without additional fine-tuning of either model).
Regarding claim 2, Zhang fails to teach but Keysers teaches: The system of claim 1, the memory further including instructions executable by the processor to: provide a set of training data as input to the first neural network and the second neural network, the set of training data including a plurality of training microstate pairs, where each respective training microstate pair of the plurality of training microstate pairs includes a first training microstate instance belonging to the first microstate space and a second training microstate instance belonging to the second microstate space; and jointly train the first neural network to learn the first mapping and train the second neural network to learn the second mapping using the set of training data, such that a difference between the first macrostate and the second macrostate is minimized for each training microstate pair of the plurality of training microstate pairs of the set of training data (¶23, For example, when the first and second ML models operate on image and text data, respectively, the first and second ML models may be trained to generate similar latent representations when an image is described (e.g., accurately described) by corresponding text, and generate dissimilar latent representations when an image is not described (e.g., described inaccurately) by corresponding text. As one example, when the image depicts a dog, the latent representation of the image should be similar to the latent representation of the text string “dog” (and other text strings that describe the image of the dog), but should be dissimilar to the latent representation of the text string “tractor” (and other text strings that do not describe the image of the dog) – here, the latent representation is the macrostate, and the microstate spaces are the image space and the text space).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to jointly train models to learn similar mappings for similar concepts as disclosed by Keysers in the system disclosed by Zhang to enable the trained models to easily adapt to new tasks (¶24, ML models trained in this manner may be configured to perform new tasks without additional data (i.e., zero-shot learning, sometimes referred to as zero-shot transfer) and/or without additional fine-tuning of either model).
Regarding claim 3, Zhang teaches: A method, comprising: applying an example microstate instance of a microstate space as input to an neural network to obtain an example macrostate of the example microstate instance (pg. 4, see figure 1 and caption, Obtain y and uncertainty from x in forward process. The posterior distribution of x from a specific y is acquired by sampling z and fixing σ using the inverse process), the neural network being one of a first neural network having learned a first mapping between a first microstate space… and a first macrostate (pg. 4, see figure 1 and caption, Obtain y and uncertainty from x in forward process – and – pg. 9, We train the P-INN)… and sampling, by the neural network, an ensemble of sampled microstate instances of the first microstate space… that correspond to the example macrostate, the neural network being an invertible neural network (pg. 6, A one-to-many mapping of the inverse process (a set of y generates multiple sets of x) is implemented by sampling z multiple times).
Zhang fails to teach: of a microstate pair… and a second neural network having learned a second mapping between a second microstate space of the microstate pair and a second macrostate… or the second microstate space
However, in the same field of endeavor, Keysers teaches: of a microstate pair… and a second neural network having learned a second mapping between a second microstate space of the microstate pair and a second macrostate… or the second microstate space (¶2, A first machine learning model and a second machine learning model, each configured to process a different type of data, may be trained using a contrastive learning process to generate similar latent representation for matched pairs of input samples of the different types of data). Zhang discusses a single model approach, with a loss function comprising multiple parts. A person of ordinary skill in the art could train two of Zhang’s P-INN models on different input data types, similar to Keyser, and add an additional contrastive loss term to the loss functions of each model. This two model approach derived from Zhang in view Keysers will be applied to all dependent claims.
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to train another network that operates on a different type of data as disclosed by Keysers in the method disclosed by Zhang to enable the trained models to easily adapt to new tasks (¶24, ML models trained in this manner may be configured to perform new tasks without additional data (i.e., zero-shot learning, sometimes referred to as zero-shot transfer) and/or without additional fine-tuning of either model).
Regarding claim 4, Zhang further teaches: The method of claim 3, further comprising: sampling, by application of the example macrostate as input to the first neural network, the ensemble of sampled microstate instances of the first microstate space that correspond to the example macrostate of the example microstate instance, the first neural network being a first invertible neural network (pg. 6, A one-to-many mapping of the inverse process (a set of y generates multiple sets of x) is implemented by sampling z multiple times).
Regarding claim 5, Zhang further teaches: The method of claim 4, further comprising: inverting the first invertible neural network (pg. 4, see figure 1 and caption, Obtain y and uncertainty from x in forward process. The posterior distribution of x from a specific y is acquired by sampling z and fixing σ using the inverse process).
Regarding claim 6, Zhang in view of Keyser further teaches: The method of claim 3, further comprising: sampling, by application of the example macrostate as input to the second neural network, the ensemble of sampled microstate instances of the second microstate space that correspond to the example macrostate of the example microstate instance, the second neural network being a second invertible neural network (Zhang, pg. 6, A one-to-many mapping of the inverse process (a set of y generates multiple sets of x) is implemented by sampling z multiple times).
Regarding claim 7, Zhang in view of Keyser further teaches: The method of claim 6, further comprising: inverting the second invertible neural network (Zhang, pg. 4, see figure 1 and caption, Obtain y and uncertainty from x in forward process. The posterior distribution of x from a specific y is acquired by sampling z and fixing σ using the inverse process).
Regarding claim 8, Zhang fails to teach but Keysers teaches: The method of claim 3, the first microstate space and the second microstate space each including observation data about a physical system, where the first microstate space corresponds with a first type of observation data about the physical system, and where the second microstate space corresponds with a second type of observation data about the physical system (¶23, For example, when the first and second ML models operate on image and text data, respectively, the first and second ML models may be trained to generate similar latent representations when an image is described (e.g., accurately described) by corresponding text, and generate dissimilar latent representations when an image is not described (e.g., described inaccurately) by corresponding text. As one example, when the image depicts a dog, the latent representation of the image should be similar to the latent representation of the text string “dog” (and other text strings that describe the image of the dog), but should be dissimilar to the latent representation of the text string “tractor” (and other text strings that do not describe the image of the dog) – here, the latent representation is the macrostate, and the microstate spaces are the image space and the text space. Dogs are part of a physical system (under BRI, the world can be interpreted as a physical system), and image and text data are different types of observation data about the world).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to train networks that operate on different types of observation data describing a physical system as disclosed by Keysers in the method disclosed by Zhang to enable the trained models to easily adapt to new tasks (¶24, ML models trained in this manner may be configured to perform new tasks without additional data (i.e., zero-shot learning, sometimes referred to as zero-shot transfer) and/or without additional fine-tuning of either model).
Regarding claim 10, Zhang fails to teach but Keysers teaches: The method of claim 3, further comprising: providing a set of training data as input to the first neural network and the second neural network, the set of training data including a plurality of training microstate pairs, where each respective training microstate pair of the plurality of training microstate pairs includes a first training microstate instance belonging to the first microstate space and a second training microstate instance belonging to the second microstate space; and jointly training the first neural network to learn the first mapping between the first microstate space and the first macrostate and training the second neural network to learn the second mapping between the second microstate space and the second macrostate using the set of training data, such that a difference between the first macrostate and the second macrostate is minimized for each training microstate pair of the plurality of training microstate pairs of the set of training data (¶23, For example, when the first and second ML models operate on image and text data, respectively, the first and second ML models may be trained to generate similar latent representations when an image is described (e.g., accurately described) by corresponding text, and generate dissimilar latent representations when an image is not described (e.g., described inaccurately) by corresponding text. As one example, when the image depicts a dog, the latent representation of the image should be similar to the latent representation of the text string “dog” (and other text strings that describe the image of the dog), but should be dissimilar to the latent representation of the text string “tractor” (and other text strings that do not describe the image of the dog) – here, the latent representation is the macrostate, and the microstate spaces are the image space and the text space).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to jointly train models to learn similar mappings for similar concepts as disclosed by Keysers in the method disclosed by Zhang to enable the trained models to easily adapt to new tasks (¶24, ML models trained in this manner may be configured to perform new tasks without additional data (i.e., zero-shot learning, sometimes referred to as zero-shot transfer) and/or without additional fine-tuning of either model).
Regarding claim 11, Zhang in view of Keysers further teaches: The method of claim 10, the ensemble of sampled microstate instances of the first microstate space following a conditional distribution where values of the ensemble of sampled first microstate instances are contingent upon the second macrostate, and the ensemble of sampled microstate instances of the second microstate space following a conditional distribution where values of the ensemble of sampled second microstate instances are contingent upon the first macrostate (Zhang, pg. 6, A one-to-many mapping of the inverse process (a set of y generates multiple sets of x) is implemented by sampling z multiple times. Define the complete distribution of results for inverse process as shown in Eq (6)
PNG
media_image1.png
73
226
media_image1.png
Greyscale
where the denominator part is the Jacobian determinant embodied by the inverse process – also refer to Zhang, pg. 4, fig. 1. In Zhang, the inverse process of the invertible network is conditioned on the “set of y” (i.e., macrostate) input into the model, and generates multiple sets of x (i.e., microstate instances). As seen in the equation, Zhang’s P-INN inherently creates microstates that are conditioned on the input macrostates. In a two model approach derived from Zhang in view of Keysers, inputting any macrostate from any training example into either of the two P-INN-like models would produce microstates conditioned on that macrostate).
Regarding claim 12, Zhang teaches: A method, comprising: providing a set of training data as input to a first neural network (pg. 9, We train the P-INN with 6 invertble blocks and the si and ti functions of the invertble block both use a 5-layer NN. All hidden layers are followed by a dropout layer with a dropout coefficient of 0.1. Train on datasets of size 50, 100 and 150 while optimizing for 3000 iterations using the Adam optimizer)… the set of training data including a plurality of training microstate… where each respective training microstate… includes a first training microstate instance belonging to a first microstate space (Zhang discusses a loss function used to train the model and involving training data – pg. 7, This term aims to narrow the difference between the inverse prediction distribution and the prior knowledge of input parameters from training samples… x∗ refers to the authentic input data – and – pg. 7, The purpose of this loss term is to make the predicted output response and true value as identical as possible, which is the main goal of forward process. The MSE loss term is defined in Eq (7)… where ŷ(i) is the forward predicted output response value, y∗(i) is the ground-truth response value these)… the first microstate space… including observation data about a physical system, where the first microstate space corresponds with a first type of observation data about the physical system (Zhang trains a P-INN to model turbine blades in relation to pressure and thermal effects – pg. 16, We first calculate the stress and strain that are only affected by the pressure while ignoring the thermal effect; then analyze the stress and strain that only has the thermal effect while ignoring the pressure; finally, the simulation model with the combined effect of the two factors is obtained – the simulation model utilize observation data about a physical system)… and… training the first neural network to learn a first mapping between the first microstate space and a first macrostate… using the set of training data (the purpose of the P-INN is to learn an invertible mapping between microstates and macrostates – see pg. 4, fig. 1. In the case of the turbine blade example, the goal is to learn a mapping between pressure/thermal model parameters and stress on the blade – see pg. 16, Here, kapp is fixed and the remaining 5 variables are taken as input. Each variable ranges from 80% to 120% of the recommended value, sampling 100 sets of input data evenly. The simulated stress is used as output)…
Zhang fails to teach: and a second neural network… pairs… pair of the plurality of training microstate pairs… and a second training microstate instance belonging to a second microstate space… and the second microstate space each… and where the second microstate space corresponds with a second type of observation data about the physical system… jointly… and training the second neural network to learn a second mapping between the second microstate space and a second macrostate…. such that a difference between the first macrostate and the second macrostate is minimized for each training microstate pair of the plurality of training microstate pairs of the set of training data.
However, in the same field of endeavor, Keysers teaches: and a second neural network… pairs… pair of the plurality of training microstate pairs… and a second training microstate instance belonging to a second microstate space… and the second microstate space each… and where the second microstate space corresponds with a second type of observation data about the physical system… jointly… and training the second neural network to learn a second mapping between the second microstate space and a second macrostate…. such that a difference between the first macrostate and the second macrostate is minimized for each training microstate pair of the plurality of training microstate pairs of the set of training data (¶23, For example, when the first and second ML models operate on image and text data, respectively, the first and second ML models may be trained to generate similar latent representations when an image is described (e.g., accurately described) by corresponding text, and generate dissimilar latent representations when an image is not described (e.g., described inaccurately) by corresponding text. As one example, when the image depicts a dog, the latent representation of the image should be similar to the latent representation of the text string “dog” (and other text strings that describe the image of the dog), but should be dissimilar to the latent representation of the text string “tractor” (and other text strings that do not describe the image of the dog) – here, the latent representation is the macrostate, and the microstate spaces are the image space and the text space, which both describe observations of a physical system (i.e., the world)). Zhang discusses a single model approach, with a loss function comprising multiple parts. A person of ordinary skill in the art could train two of Zhang’s P-INN models on different input data types, similar to Keyser, and add an additional contrastive loss term to the loss functions of each model. Keyser also mentions in ¶48: Although text-image pairs are used herein as examples, it is to be understood that encoders 316 and 318 may alternatively represent machine learning models configured to process any combination of two different types of data for which training samples are available. This two model approach derived from Zhang in view Keysers will be applied to all dependent claims.
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to jointly train models to learn similar mappings for similar paired concepts as disclosed by Keysers in the method disclosed by Zhang to enable the trained models to easily adapt to new tasks (¶24, ML models trained in this manner may be configured to perform new tasks without additional data (i.e., zero-shot learning, sometimes referred to as zero-shot transfer) and/or without additional fine-tuning of either model).
Regarding claim 15, Zhang in view of Keysers further teaches: The method of claim 12, the first microstate space following a conditional distribution where values of an ensemble of first microstate instances of the first microstate space are contingent upon the second macrostate, and the second microstate space following a conditional distribution where values of an ensemble of second microstate instances of the second microstate space are contingent upon the first macrostate (Zhang, pg. 6, A one-to-many mapping of the inverse process (a set of y generates multiple sets of x) is implemented by sampling z multiple times. Define the complete distribution of results for inverse process as shown in Eq (6)
PNG
media_image1.png
73
226
media_image1.png
Greyscale
where the denominator part is the Jacobian determinant embodied by the inverse process – also refer to Zhang, pg. 4, fig. 1. In Zhang, the inverse process of the invertible network is conditioned on the “set of y” (i.e., macrostate) input into the model, and generates multiple sets of x (i.e., microstate instances). As seen in the equation, Zhang’s P-INN inherently creates microstates that are conditioned on the input macrostates. In a two model approach derived from Zhang in view of Keysers, inputting any macrostate from any training example into either of the two P-INN-like models would produce microstates conditioned on that macrostate).
Regarding claim 16, Zhang in view of Keysers further teaches: The method of claim 12, further comprising: applying an example microstate instance of the first microstate space or the second microstate space as input to the first neural network or the second neural network (Zhang, pg. 4, fig. 1, see the definition of the forward process – input higher dimensional x); determining an example macrostate of the example microstate instance using the first neural network or the second neural network (Zhang, pg. 4, fig. 1, see the definition of the forward process – from x, derive y (i.e., a macrostate) among other values); and sampling an ensemble of sampled microstate instances of the first microstate space or of the second microstate space that correspond to the example macrostate (Zhang, pg. 6, The inverse process takes response y, predefined σ, and the sampled z as input, and derives the x… A one-to-many mapping of the inverse process (a set of y generates multiple sets of x) is implemented by sampling z multiple times) [[note: claim 16 can be read as: The method of claim 12, further comprising: applying an example microstate instance of the first microstate space [[]].
Regarding claim 17, Zhang further teaches: The method of claim 16, further comprising: sampling, by application of the example macrostate as input to the first neural network, the ensemble of sampled microstate instances of the first microstate space that correspond to the example macrostate, the first neural network being a first invertible neural network (pg. 6, The inverse process takes response y, predefined σ, and the sampled z as input, and derives the x… A one-to-many mapping of the inverse process (a set of y generates multiple sets of x) is implemented by sampling z multiple times).
Regarding claim 18, Zhang further teaches: The method of claim 17, further comprising: inverting the first invertible neural network (pg. 6, The inverse process takes response y, predefined σ, and the sampled z as input, and derives the x… A one-to-many mapping of the inverse process (a set of y generates multiple sets of x) is implemented by sampling z multiple times).
Regarding claim 19, Zhang in view of Keysers further teaches: The method of claim 16, further comprising: sampling, by application of the example macrostate as input to the second neural network, the ensemble of sampled microstate instances of the second microstate space that correspond to the example macrostate, the second neural network being a second invertible neural network (Zhang, pg. 6, The inverse process takes response y, predefined σ, and the sampled z as input, and derives the x… A one-to-many mapping of the inverse process (a set of y generates multiple sets of x) is implemented by sampling z multiple times).
Regarding claim 20, Zhang in view of Keysers further teaches: The method of claim 19, further comprising: inverting the second invertible neural network (Zhang, pg. 6, The inverse process takes response y, predefined σ, and the sampled z as input, and derives the x… A one-to-many mapping of the inverse process (a set of y generates multiple sets of x) is implemented by sampling z multiple times).
Claim(s) 9 and 13 is/are rejected under 35 U.S.C. 103 as being unpatentable over Zhang in view of Keysers as applied to claims 3 and 12 above, and further in view of Liu et al. (US 20180247201 A1), herein Liu.
Regarding claim 9, Zhang in view of Keysers fails to explicitly teach: The method of claim 3, the microstate pair following a joint distribution that relates the first microstate space to the second microstate space.
However, in the same field of endeavor, Liu teaches: the microstate pair following a joint distribution that relates the first microstate space to the second microstate space (¶24, FIG. 1B illustrates correlated image pairs 160 for supervised image-to-image translation training and uncorrelated images 165 for unsupervised image-to-image translation training, in accordance with one embodiment. For supervised training, paired of correlated images (x1, x2) drawn from a joint distribution PX1, X2 (x1, x2) of the domains x1 and x2 are available – Liu’s process of mapping an image to a latent space, and back to an image as described in fig. 2A for example, is similar to the invertible mapping process described by Zhang).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use paired training data that follows a joint distribution as disclosed by Liu in the method disclosed by Zhang in view of Keysers to simplify the distribution learning process (Liu describes that supervised training using correlated image pairs is traditionally used: ¶4, Traditionally, supervised training is used to train the neural network model. The supervised training requires a training dataset with image pairs that include an image in the first domain that is perfectly correlated with an image in the second domain – and indirectly describes supervised learning as easier – ¶24, Due to lack of correlated images, the unsupervised training of an image-to-image translation system is considered more difficult).
Regarding claim 13, Zhang fails to teach: The method of claim 12, where jointly training the first neural network and the second neural network includes iteratively determining parameters of the first neural network and the second neural network that minimize a loss function incorporating a prediction loss between results of the first mapping and the second mapping for each training microstate pair of the plurality of training microstate pairs…
However, in the same field of endeavor, Keysers teaches: where jointly training the first neural network and the second neural network includes iteratively determining parameters of the first neural network and the second neural network (¶50, training system 300 may be configured to perform the multimodal contrastive training by iteratively updating parameters of text encoder 318 until text encoder 318 learns to generate text latent representations that are coordinated with the image latent representations generated by pretrained image encoder 316 – Keysers discusses locking the weights of one model while adjusting the weights of the other, but also describes adjusting parameters in both models simultaneously: ¶53, some parameters of image encoder 316 may be held fixed (i.e., locked) by training system 300 for the duration of training, and other parameters of image encoder 316 may be modifiable during training) that minimize a loss function incorporating a prediction loss between results of the first mapping and the second mapping for each training microstate pair of the plurality of training microstate pairs (¶58, Contrastive loss function 324 may be configured to generate loss value 326 based on comparing the image latent representations generated by pretrained image encoder 316 to the text latent representations generated by text encoder 318. Specifically, loss value 326 may be based on comparing the respective latent representations of matched image-text training samples, and/or unmatched image-text training samples. Contrastive loss function 324 may be configured to, for matched image-text training samples, incentivize text encoder 318 to generate text latent representations that are similar to corresponding image latent representations generated by pretrained image encoder 316)…
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to jointly train models to learn similar mappings for similar concepts as disclosed by Keysers in the system disclosed by Zhang to enable the trained models to easily adapt to new tasks (¶24, ML models trained in this manner may be configured to perform new tasks without additional data (i.e., zero-shot learning, sometimes referred to as zero-shot transfer) and/or without additional fine-tuning of either model).
Zhang in view of Keysers fails to teach: where the first microstate space is related to the second microstate space by a joint distribution.
However, in the same field of endeavor, Liu teaches: where the first microstate space is related to the second microstate space by a joint distribution (¶24, FIG. 1B illustrates correlated image pairs 160 for supervised image-to-image translation training and uncorrelated images 165 for unsupervised image-to-image translation training, in accordance with one embodiment. For supervised training, paired of correlated images (x1, x2) drawn from a joint distribution PX1, X2 (x1, x2) of the domains x1 and x2 are available – Liu’s process of mapping an image to a latent space, and back to an image as described in fig. 2A, for example, is similar to the invertible mapping process described by Zhang).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use paired training data that follows a joint distribution as disclosed by Liu in the method disclosed by Zhang in view of Keysers to simplify the distribution learning process (Liu describes that supervised training using correlated image pairs is traditionally used: ¶4, Traditionally, supervised training is used to train the neural network model. The supervised training requires a training dataset with image pairs that include an image in the first domain that is perfectly correlated with an image in the second domain – and indirectly describes supervised learning as easier – ¶24, Due to lack of correlated images, the unsupervised training of an image-to-image translation system is considered more difficult).
Allowable Subject Matter
Claim 14 is objected to as being dependent upon a rejected base claim, but would be allowable over prior art if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Xu et al. (US 20220082718 A1), discusses how the Jacobian should never vanish to ensure transformations are invertible.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to HARRISON CHAN YOUNG KIM whose telephone number is (571)272-0713. The examiner can normally be reached Monday - Friday 10:00 am - 6:00 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, CESAR PAULA can be reached at (571) 272-4128. 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.
/HARRISON C KIM/ Examiner, Art Unit 2145
/CESAR B PAULA/Supervisory Patent Examiner, Art Unit 2145
