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 .
Status of Claims
Claims 1-9 are pending and are examined herein.
Claims 1, 4-6, and 8 are objected to.
Claims 1-5 are rejected under 35 U.S.C. 101.
Claims 1-9 are rejected under 35 U.S.C. 112(b).
Claims 1-9 are rejected under 35 U.S.C. 103.
Claim Objections
Claims 1, 4-6, and 8 are objected to because of the following informalities:
In claims 1, 6, and 8, there should be an “and” in the penultimate paragraph after the semicolon.
In claims 4 and 5, the claims should be modified to include the definition of INx,y, which is the input plane according to the specification.
In claim 4, the claim should be modified to include the definition of OUTa,b, which is the output.
In claim 5, the claim should be modified to include the definition of OUTx,y, which is the output.
Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-9 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
The term “complex” in claim 1 is a relative term which renders the claim indefinite. The term “complex” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. The use of the term “complex” in claim 1 renders the term “complex optical fields” indefinite, as it is unclear which optical fields would be considered “complex.” For purposes of examination, any optical field will be interpreted as a complex optical field.
The term “akin to activation functions” in claim 1 is a relative term which renders the claim indefinite. The term “akin to activation functions” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree of likeness, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. The use of the term “akin to activation functions” in claim 1 renders the term “noise models” indefinite, as it is unclear which noise models would be considered “akin to activation functions.” For purposes of examination, any noise model will be interpreted as a noise model akin to activation functions.
Dependent claims 2-9 fail to resolve the issues and are rejected with the same rationales, mutatis mutandis.
Claim 2 recites the limitation "the optical system components." There is insufficient antecedent basis for this limitation in the claim. As there is insufficient antecedent basis, it is unclear if the input layer and detection layer comprise optical system components, or if the limitation refers to the optical components that represent the modulator layers as recited in claim 1. For purposes of examination, each layer will be interpreted as comprising an optical system component.
Claim 2 recites the limitation "the optical system." There is insufficient antecedent basis for this limitation in the claim. As there is insufficient antecedent basis, it is unclear if this limitation refers to “optical systems” in the preamble of claim 1, or if there is another optical system. For purposes of examination, each layer will be interpreted as comprising an optical system component.
The term “precise” in claim 2 is a relative term which renders the claim indefinite. The term “precise” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. The use of the term “precise” in claim 2 renders the term “precise measurement, alignment, modeling, and description of the optical system” indefinite.
Regarding claim 4, the term Wx,y,a,b is undefined. As it is undefined in the claim, it is unclear whether Wx,y,a,b is a separate matrix from the transfer matrix of the propagation layer, or if it is the same matrix. Note that, as claim 4 depends only on claim 1, the definition of Wx,y,a,b from claim 3 is not included in claim 4. For purposes of examination, Wx,y,a,b will be interpreted as transfer matrix that defines the propagation layer, as presented in claim 3.
Regarding claim 5, the term Wx,y,a,b is undefined. As it is undefined in the claim, it is unclear whether Wx,y,a,b is a separate matrix from the transfer matrix of the propagation layer, or if it is the same matrix. Note that, as claim 5 depends only on claim 1, the definition of Wx,y,a,b from claim 3 is not included in claim 5. For purposes of examination, Wx,y,a,b will be interpreted as transfer matrix that defines the propagation layer, as presented in claim 3.
Regarding claim 5, the phrase "such as" renders the claim indefinite because it is unclear whether the limitations following the phrase are part of the claimed invention. See MPEP § 2173.05(d).
The term “accurate” in claim 8 is a relative term which renders the claim indefinite. The term “accurate” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. The use of the term “accurate” in claim 8 renders the term “accurate measurement and description of the optical system” indefinite.
Dependent claim 9 fails to resolve the issue and is rejected with the same rationale, mutatis mutandis.
The term “accurate” in claim 9 is a relative term which renders the claim indefinite. The term “accurate” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. The use of the term “accurate” in claim 9 renders the term “accurate description of the optical system” indefinite.
The term “precise” in claim 9 is a relative term which renders the claim indefinite. The term “precise” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. The use of the term “precise” in claim 9 renders the term “precise simulation” indefinite.
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 non-statutory subject matter.
The claim(s) does/do not fall within at least one of the four categories of patent eligible subject matter because they are directed to “software per se.” The invention is claimed as a product, an “Optical Intrinsic Neural Network.” However, in the broadest reasonable interpretation, the claims do not recite a tangible structural component. Each recited component may be represented by a software component. See MPEP 2106.03(I).
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-5 is/are rejected under 35 U.S.C. 103 as being obvious over US 2021/0072784 A1, hereinafter, “Lin” and Zibo Hu et al., "Black-box simulation method: train the optical model from output," April 14, 2023, SPIE, Physics and Simulation of Optoelectronic Devices XXXI, Proceedings of SPIE Vol. 12415, hereinafter “Hu.”
The applied reference, Hu, has a common inventor with the instant application. Based upon the earlier effectively filed date of the reference, it constitutes prior art under 35 U.S.C. 102(a)(1). As the reference was published on 04/14/2023, which is before the one-year grace period of the effective filing date of the application, 08/12/2024, the reference is not excepted as prior art under 35 U.S.C. 102(b)(1).
Regarding claim 1, Lin teaches
An Optical Intrinsic Neural Network (OINN) . . . comprising: ([0021] states "FIG. 1 illustrates an example integrated photonic device 100 that comprises an ONN [Optical Neural Network] 102 that includes one or more layers 104 having multiple optical signal inputs 106 and multiple optical signal outputs 108." The ONN is interpreted as the optical intrinsic neural network.)
an Input Layer configured to receive incoming light; ([0022] states "Within the ONN 102, a laser diode array (LDA) 110 together with optical modulators 112 (hereinafter referred to as “modulator 112”) provides optical input to a first layer 105." Therefore, the first layer is interpreted as the input layer.)
a plurality of Modulator Layers, each representing an optical component and defined by complex optical fields and parameter matrices with intrinsic physical meanings; ([0022] states "In this example, light signals are sent from layer 1 105, to layer 2 104, and then to layer 3 107. Each layer is made up of an optical unitary matrix multiplier (that may include a plurality of optical unitary matrix multipliers) and non-linear optical devices (e.g., nonlinear optical amplifiers 124 described below)." The optical unitary matrix multipliers are interpreted as the modulator layers. [0023] states "As shown, layer 2 104 includes three optical unitary matrix multipliers 118, 120, 122 that are composed of a plurality of optical unitary matrices (not shown), coupled together as described with respect to FIG. 7 below." The optical unitary matrices are interpreted as the parameter matrices. As the optical unitary matrix multipliers are a physical optical component that has optical signals (light) moving through them, they are defined by complex optical fields. As the optical neural network is a physical neural network, the parameter matrices in the neural network have intrinsic physical meanings.)
a plurality of Propagation Layers interspersed between said Modulator Layers, each defined by complex optical fields, parameter matrices, and incorporating noise models akin to activation functions in neural networks; ([0024] states "Nonlinear optical amplifiers 124 may be needed to be coupled to the optical unitary matrix multiplier 122 due to the linear nature of the optical signal processing from the optical unitary matrix multipliers 118, 120, 122." The nonlinear optical amplifiers are interpreted as the propagation layers. As the nonlinear optical amplifiers are a physical optical component that has optical signals (light) moving through them, they are defined by complex optical fields. [0026] states "First, the optical nonlinear activation may need active feedback control to emulate the arbitrary layers matrices and to classify and predict performance. Examples of active control are bias predict performance. Examples of active control are bias current, voltage and/or phase tuning operation for activation functions in optical amplifying, attenuating and saturating." [0029] states "Gain medium 238 may be bonded on the silicon photonics waveguide 234 to provide nonlinear output power photonics waveguide 234 to provide nonlinear output power change with optical input power, in bias current vs voltage operation for activation functions in optical amplifying, attenuating and saturating." Therefore, the arbitrary layer matrices, interpreted as the parameter matrices defining the propagation layers, are emulated. [0026] states "The equation
I
o
u
t
=
f
I
i
n
e
t
∆
ϕ
on the output of 126 shown in FIG. 1 defines the overall optical signal input to optical signal output nonlinear activation function, where f is the optical intensity function of nonlinear optical device 128 as a function of optical signal input power
I
i
n
; and
∆
ϕ
is the phase changes from optical signal input to optical signal output generated by the non-linear optical device 128." Therefore, the nonlinear activation function provided by the propagation layers are interpreted as the noise model. As the nonlinear optical amplifiers are a physical optical component that has optical signals (light) moving through them, they are defined by complex optical fields.)
a Detection Layer configured to observe optical intensity and defined by complex optical fields and parameter matrices, including noise models; ([0022] states "A photodetector array 114 will receive optical output from the third layer 107, and convert that output into digital signals." The photodetector array is interpreted as the detection layer. As it receives the optical output, it observes the optical intensity. As the optical neural network is a physical neural network, the parameter matrices in the neural network have intrinsic physical meanings. As the photodetector array (PDA) is a physical optical component that has optical signals (light) moving through it, it is defined by complex optical fields. The third layer parameter matrices are interpreted as the parameter matrices that define the detection layer, as the detection layer will receive the results of the third layer. Photodetector arrays inherently contain noise, meaning that noise models are included.)
an Output Layer configured to provide data received by a computer as system output. ([0022] states "A photodetector array 114 will receive optical output from the third layer 107, and convert that output into digital signals." As the output is converted to digital output, one of ordinary skill in the art would reason that the data is received by a computer.)
Lin does not appear to explicitly teach
[an optical neural network] for measuring, aligning, modeling, and describing optical systems,
However, Hu—directed to analogous art—teaches
[an optical neural network] for measuring, aligning, modeling, and describing optical systems, (The abstract states "Accurate simulation is a critical requirement for modern optical systems, but precise theoretical modeling can be challenging due to various factors such as misalignment, theoretical approximations, and instrument errors. This paper presents the black-box simulation method, which addresses these limitations by training the optical system model using the optical field outputs. This approach leads to more accurate simulations, making it possible to effectively train optical Neural Network systems for improved performance." Therefore, as the optical neural network is trained to model the optical system, the optical neural network is for measuring, aligning, modeling, and describing optical systems.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of this application to combine the teachings of Lin and Hu, because as Hu states on pages 1-2, “Previous studies attempted to train the system through simulations with the aim of building an optical system that perfectly matched the design, but this resulted in a 20% reduction in classification accuracy compared to the simulation. Despite implementing compensation algorithms, the CIFAR-10 classification accuracy still dropped by 10%. This is due to the slow optoelectronic domain crossing, which makes it challenging to train the neural network parameters on actual optical settings.83–94 Our approach, the black-box simulation method, solves this problem by scanning an actual optical system to build an accurate simulation model, rather than trying to build an accurate optical system from a theoretical design. This accurate simulation approach makes it easier to scale the optical system by treating the scanning system as an optical module and cascading multiple modules." Additionally, as stated by Hu on page 1, "To mitigate this issue, it is crucial to train the optical neural networks on the actual experimental setup rather than relying solely on simulated training."
Regarding claim 2, the rejection of claim 1 is incorporated herein. Lin teaches
wherein the Input Layer, Modulator Layers, and Detection Layer are configured according to established physical relationships of the optical system components, thereby ensuring precise measurement, alignment, modeling, and description of the optical system. ([0022] states "Within the ONN 102, a laser diode array (LDA) 110 together with optical modulators 112 (hereinafter referred to as “modulator 112”) provides optical input to a first layer 105. A photodetector array 114 will receive optical output from the third layer 107, and convert that output into digital signals. In this example, light signals are sent from layer 1 105, to layer 2 104, and then to layer 3 107. Each layer is made up of an optical unitary matrix multiplier (that may include a plurality of optical unitary matrix multipliers) and non-linear optical devices (e.g., nonlinear optical amplifiers 124 described below). In embodiments, the ONN 102 including array (LDA) 110, modulator 112, multiple layers 105, 104, 107, and PDA 114 can be implemented in a heterogeneously integrated photonics circuit, such as a single silicon photonics die or single semiconductor substrate 150." As each layer is configured for optical input/optical output to the next layer, they are configured according to established physical relationships of the optical system components
Regarding claim 3, the rejection of claim 1 is incorporated herein. Lin teaches
wherein the Propagation Layer is defined by a transfer matrix Wx,y,a,b, such that the output OUTa,b is calculated as:
OUTa,b=∑ INx,y ⋅Wx,y,a,b
Where x, y are the coordinates of the input plane, and a, b are the coordinates of the output plane. (Fig. 1 shows that the output of the propagation layer, which includes optical unitary matrix *multipliers 118, 120, and 122, is calculated by
M
=
U
∑
V
+
. Therefore,
V
+
is transfer matrix INx,y, and U is Wx,y,a,b. [0050]-[0051] state "Specifically, FIG. 7 below provides example configurations of the layer 104 and specifically matrices 118, 120, and 122 of FIG. 1. As described in reference to FIG. 1, the nonlinear optical amplifier 124 (or device 128) may be used in connection with the layer 104, described in detail in reference to FIG. 7. [0051] FIG. 7 illustrates an example matrix multiplier that includes a plurality of 2x2 unitary directional optical matrices and an optical unitary matrix that includes a plurality of 2x2 unitary multi-mode interference (MMI) optical couplers, in accordance with other embodiments of the present disclosure." Therefore, as the unitary directional optical matrices are 2x2, one of ordinary skill in the art would be able to reason that the output at coordinates a,b would be calculated based on the input at coordinates x,y multiplied by the transfer matrix at coordinates a,b,x,y.)
Regarding claim 4, the rejection of claim 1 is incorporated herein.
wherein the Modulator Layer incorporates nonlinear relationships specific to different modulators, defined by the equation:
OUTa,b = 𝑓 ( ∑ 𝑔 (INx,y ⋅Wx,y,a,b ))
where 𝑓() and 𝑔() are nonlinear functions, x, y are the coordinates of the input plane, and a, b are the coordinates of the output plane. ([0026] states "The equation
I
o
u
t
=
f
I
i
n
e
t
∆
ϕ
on the output of 126 shown in FIG. 1 defines the overall optical signal input to optical signal output nonlinear activation function, where f is the optical intensity function of nonlinear optical device 128 as a function of optical signal input power
I
i
n
; and
∆
ϕ
is the phase changes from optical signal input to optical signal output generated by the non-linear optical device 128."
f
e
t
∆
ϕ
is interpreted as f(). [0024] states "The optical signal, including noise added to the optical signal, may be linearly increased during operation of the ONN 102, and may result in a final signal intensity from the Un optical unitary matrix multiplier 122 that is too high." The noise added to the optical signal of
I
i
n
is interpreted as g(), which one of ordinary skill in the art would be non-linear, such as in the case of shot noise, which is inherent in an optical system. As the optical unitary matrix multiplier is coupled to the nonlinear optical device,
I
i
n
is the result of the matrix multiplication units 118, 120, and 120 (propagation layer). Fig. 1 shows that the output of the propagation layer, which includes optical unitary matrix multipliers 118, 120, and 122, is calculated by
M
=
U
∑
V
+
. Therefore,
V
+
is transfer matrix Wx,y,a,b, and U is INx,y. [0050]-[0051] state "Specifically, FIG. 7 below provides example configurations of the layer 104 and specifically matrices 118, 120, and 122 of FIG. 1. As described in reference to FIG. 1, the nonlinear optical amplifier 124 (or device 128) may be used in connection with the layer 104, described in detail in reference to FIG. 7. [0051] FIG. 7 illustrates an example matrix multiplier that includes a plurality of 2x2 unitary directional optical matrices and an optical unitary matrix that includes a plurality of 2x2 unitary multi-mode interference (MMI) optical couplers, in accordance with other embodiments of the present disclosure." Therefore, as the unitary directional optical matrices are 2x2, one of ordinary skill in the art would be able to reason that the output at coordinates a,b would be calculated based on the input at coordinates x,y and the transfer matrix at coordinates a,b,x,y.)
Regarding claim 5, the rejection of claim 1 is incorporated herein. Lin teaches
wherein the Detection Layer includes noise models such as shot noise and thermal noise, defined by the equation:
OUTx,y = 𝑓 ( INx,y ⋅Wx,y,a,b )
where 𝑓() includes noise and nonlinear functions, x, y are the coordinates of the input plane, and a, b are the coordinates of the output plane. (As the photodetector array is interpreted as the detection layer, it inherently includes noise models such as shot noise and thermal noise, which is interpreted as part of f(). As it detects the output of the previous layer, the calculation of the propagation layer is included in the Fig. 1 shows that the output of the propagation layer, which includes optical unitary matrix multipliers 118, 120, and 122, is calculated by
M
=
U
∑
V
+
. Therefore,
V
+
is transfer matrix Wx,y,a,b, and U is INx,y. [0050]-[0051] state "Specifically, FIG. 7 below provides example configurations of the layer 104 and specifically matrices 118, 120, and 122 of FIG. 1. As described in reference to FIG. 1, the nonlinear optical amplifier 124 (or device 128) may be used in connection with the layer 104, described in detail in reference to FIG. 7. [0051] FIG. 7 illustrates an example matrix multiplier that includes a plurality of 2x2 unitary directional optical matrices and an optical unitary matrix that includes a plurality of 2x2 unitary multi-mode interference (MMI) optical couplers, in accordance with other embodiments of the present disclosure." Therefore, as the unitary directional optical matrices are 2x2, one of ordinary skill in the art would be able to reason that the output at coordinates a,b would be calculated based on the input at coordinates x,y and the transfer matrix at coordinates a,b,x,y. The parts of the propagation layer not represented in INx,y ⋅Wx,y,a,b is also interpreted as part of f(). As the detection layer also receives the output of the nonlinear/modulator layer, f() also includes the nonlinear function implemented by the modulator layer.)
Claim 6 is/are rejected under 35 U.S.C. 103 as being obvious over Lin and Hu as applied to claim 1 above, and further in view of US 2024/0232617 A1, hereinafter “Duan.”
Regarding claim 6, the rejection of claim 1 is incorporated herein. The combination of Lin and Hu teaches
the Optical Intrinsic Neural Network (OINN) of Claim 1 (See explanation of rejection of claim 1.)
The combination of Lin and Hu does not appear to explicitly teach
A method for training, comprising:
collecting a training dataset consisting of input-output pairs from the optical system;
configuring the OINN with corresponding Input, Modulator, Propagation, and Detection Layers;
training the OINN to minimize the loss between predicted and actual outputs, thereby determining intrinsic values approximating true physical quantities.
However, Duan—directed to analogous art—teaches
A method for training, comprising: ([0005] states "The present application provides a dual adaptive training method of photonic neural networks (PNN) and associated components, which solves a problem that the network training method in the prior art cannot adapt to continuously accumulated systematic errors in PNN, allows the PNN physical model to adapt to significant systematic errors during the training and the PNN physical model to maintain high performance when being deployed in the PNN physical system.")
collecting a training dataset consisting of input-output pairs from the optical system; ([0054] states "In order to determine the measurement values of the PNN physical system and the measurement values of the PNN numerical model, in the present embodiment, each training sample is optically encoded to obtain I, and I is input to the PNN physical system composed of N-layer networks to perform forward propagation and the final output values PN and internal states {P1, P2,..., P N_1} are obtained." The input of the PNN physical system and the final output values are interpreted as the training dataset consisting of input-output pairs, and the PNN physical system is interpreted as the optical system.)
configuring the OINN with corresponding Input, Modulator, Propagation, and Detection Layers; (Fig. 8 and Fig. 9 shows that the numerical model reflects the physical system, including each layer included in the physical optical neural network.)
training the OINN to minimize the loss between predicted and actual outputs, thereby determining intrinsic values approximating true physical quantities. ([0055] states "As an embodiment, determining the similarity loss function based on comparison results between the measurement values of the PNN physical system and the measurement values of the PNN numerical model includes: determining, based on a comparison result between the internal states and the final output values of the PNN physical system and the internal states and final output values of the PNN numerical model, that the similarity loss function
L
s
in a unitary optimization mode is." [0060] states "In an embodiment, optimizing and updating parameters of the PNN numerical model based on the similarity loss function and the task loss function for in situ training of the PNN physical model includes: minimizing the similarity loss function
L
s
in a unitary optimization mode to update parameters of the systematic error prediction network model." Therefore, as the similarity loss function is minimized, the intrinsic errors of the physical system can be approximated by the PNN numerical model.)
It would have been obvious to a person having ordinary skill in the art before the effective filing date of this application to combine the teachings of Lin and Hu with the teachings of Duan because, as stated by Duan in [0049], "To solve the problems in the prior art, the present application provides a dual adaptive training (DAT) method for large-scale PNN in which a dual adaptive backpropagation training process is constructed to end-to-end optimize the PNN physical models and enable the PNN physical models to adapt to significant systematic errors without additional hardware configurations for generation and propagation of backward optical fields."
Claim 7 is/are rejected under 35 U.S.C. 103 as being obvious over Lin and Hu as applied to claim 1 above, further in view of Duan as applied to claim 6 above, and further in view of Humayun Kabir et al., “Neural Network Inverse Modeling and Applications to Microwave Filter Design,” April 2008, IEEE, IEEE Transactions on Microwave Theory and Techniques, Vol. 56, No. 4, hereinafter “Kabir.”
Regarding claim 7, the rejection of claim 6 is incorporated herein. The combination of Lin, Hu, and Duan do not appear to explicitly teach wherein the training dataset is specially configured to ensure each output has a unique input to facilitate convergence of the OINN to a low-loss state.
However, Kabir—directed to analogous art—teaches
wherein the training dataset is specially configured to ensure each output has a unique input to facilitate convergence of the OINN to a low-loss state. (Page 869 states "If two different input values in the forward model lead to the same value of output, then a contradiction arises in the training data of the inverse model because the single input value in the inverse model has two different output values. Since we cannot train the neural network inverse model to match two different output values simultaneously, the training error cannot be reduced to a small value. As a result, the trained inverse model will not be accurate. For this reason, it is important to detect the existence of multivalued solutions, which creates contradictions in training data." Page 870 states "If existence of multivalued solutions is detected in training data, we perform data preprocessing to divide the data into different groups such that the data in each group do not have the problem of multivalued solutions." One group not containing multivalued solutions is interpreted as the training dataset, as each group is later used to train one model.)
It would have been obvious to a person having ordinary skill in the art before the effective filing date of this application to combine the teachings of Lin, Hu, and Duan with the teachings of Kuan because, as Kuan states on page 869, "If two different input values in the forward model lead to the same value of output, then a contradiction arises in the training data of the inverse model because the single input value in the inverse model has two different output values. Since we cannot train the neural network inverse model to match two different output values simultaneously, the training error cannot be reduced to a small value. As a result, the trained inverse model will not be accurate. For this reason, it is important to detect the existence of multivalued solutions, which creates contradictions in training data."
Claims 8 and 9 is/are rejected under 35 U.S.C. 103 as being obvious over Lin and Hu as applied to claim 1 above, and further in view of Duan and Kabir.
Regarding claim 8, the rejection of claim 1 is incorporated herein. The combination of Lin and Hu teach
the Optical Intrinsic Neural Network (OINN) of Claim 1, comprising: (See rejection of claim 1.)
The combination of Lin and Hu does not appear to teach
An application process for using [an optical neural network] comprising:
collecting unique input-output training sets;
constructing the OINN corresponding to specific optical components;
training the OINN to achieve the lowest loss;
extracting the parameters of the trained OINN to achieve accurate measurement and description of the optical system.
However, Duan—directed to analogous art—teaches
An application process for using [an optical neural network] comprising: ([0049] states "To solve the problems in the prior art, the present application provides a dual adaptive training (DAT) method for large-scale PNN in which a dual adaptive backpropagation training process is constructed to end-to-end optimize the PNN physical models and enable the PNN physical models to adapt to significant systematic errors without additional hardware configurations for generation and propagation of backward optical fields.")
collecting … input-output training sets; ([0054] states "In order to determine the measurement values of the PNN physical system and the measurement values of the PNN numerical model, in the present embodiment, each training sample is optically encoded to obtain I, and I is input to the PNN physical system composed of N-layer networks to perform forward propagation and the final output values PN and internal states {P1, P2,..., P N_1} are obtained." The input of the PNN physical system and the final output values are interpreted as the training dataset consisting of input-output pairs, and the PNN physical system is interpreted as the optical system.)
constructing the OINN corresponding to specific optical components; (Fig. 8 and Fig. 9 shows that the numerical model reflects the physical system, including each layer included in the physical optical neural network, which correspond to specific optical components in the system.)
training the OINN to achieve the lowest loss; ([0055] states "As an embodiment, determining the similarity loss function based on comparison results between the measurement values of the PNN physical system and the measurement values of the PNN numerical model includes: determining, based on a comparison result between the internal states and the final output values of the PNN physical system and the internal states and final output values of the PNN numerical model, that the similarity loss function
L
s
in a unitary optimization mode is." [0060] states "In an embodiment, optimizing and updating parameters of the PNN numerical model based on the similarity loss function and the task loss function for in situ training of the PNN physical model includes: minimizing the similarity loss function
L
s
in a unitary optimization mode to update parameters of the systematic error prediction network model." Therefore, as the similarity loss function is minimized, the intrinsic errors of the physical system can be approximated by the PNN numerical model.)
extracting the parameters of the trained OINN to achieve accurate measurement and description of the optical system. ([0054] states "Different from the physical system, it is easy to extract the complex optical fields from the numerical model. If the systematic errors can be perfectly characterized with SEPNs, it can be known that
S
n
2
=
P
n
. Therefore, the measurement values of the PNN physical system and the measurement values of the PNN numerical model can be accurately obtained using the method of the present embodiment." As the SEPN (part of the numerical optical neural network) characterizes the systematic errors when the numerical optical neural network is trained, the parameters of the SEPN are extracted to obtain the measurement of the optical system (PNN physical system).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of this application to combine the teachings of Lin and Hu with the teachings of Duan because, as stated by Duan in [0049], "To solve the problems in the prior art, the present application provides a dual adaptive training (DAT) method for large-scale PNN in which a dual adaptive backpropagation training process is constructed to end-to-end optimize the PNN physical models and enable the PNN physical models to adapt to significant systematic errors without additional hardware configurations for generation and propagation of backward optical fields."
The combination of Lin, Hu, and Duan does not appear to explicitly teach
[collecting] unique [training sets]
However, Kabir—directed to analogous art—teaches
[collecting] unique [training sets] (Page 869 states "If two different input values in the forward model lead to the same value of output, then a contradiction arises in the training data of the inverse model because the single input value in the inverse model has two different output values. Since we cannot train the neural network inverse model to match two different output values simultaneously, the training error cannot be reduced to a small value. As a result, the trained inverse model will not be accurate. For this reason, it is important to detect the existence of multivalued solutions, which creates contradictions in training data." Page 870 states "If existence of multivalued solutions is detected in training data, we perform data preprocessing to divide the data into different groups such that the data in each group do not have the problem of multivalued solutions." The groups are interpreted as the unique training sets.)
It would have been obvious to a person having ordinary skill in the art before the effective filing date of this application to combine the teachings of Lin, Hu, and Duan with the teachings of Kuan because, as Kuan states on page 869, "If two different input values in the forward model lead to the same value of output, then a contradiction arises in the training data of the inverse model because the single input value in the inverse model has two different output values. Since we cannot train the neural network inverse model to match two different output values simultaneously, the training error cannot be reduced to a small value. As a result, the trained inverse model will not be accurate. For this reason, it is important to detect the existence of multivalued solutions, which creates contradictions in training data."
Regarding claim 9, the rejection of claim 8 is incorporated herein. The combination of Lin and Hu does not appear to explicitly teach
wherein the trained OINN provides an accurate description of the optical system, enabling precise simulation and future optical system design. ([0049] states "In an embodiment, to precisely model the PNN physical system, a PNN numerical model constructed in a digital computer includes a PNN physical model and a systematic error prediction network (SEPN) model, to respectively model a photonic computing process and inherent systematic errors and a task-similarity loss function joint optimization approach based on the dual backpropagation training is developed." [0049] further states "After training, the PNN physical model is directly deployed on the PNN physical system and can adapt to significant systematic errors from various sources. On this base, the DAT therefore supports large-scale PNN training and mitigates the requirement of high-precision fabrication and system configurations.")
It would have been obvious to a person having ordinary skill in the art before the effective filing date of this application to combine the teachings of Lin and Hu with the teachings of Duan for the reasons given above in regards to claim 8.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JESSICA THUY PHAM whose telephone number is (571)272-2605. The examiner can normally be reached Monday - Friday, 9 A.M. - 5:00 P.M..
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, Li Zhen can be reached at (571) 272-3768. 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.
/J.T.P./Examiner, Art Unit 2121
/Li B. Zhen/Supervisory Patent Examiner, Art Unit 2121
