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 .
Response to Amendment
The amendments to the claims, in the submission dated 08/14/2026, are acknowledged and accepted. Claims 1 5-8, 11, 13 and 15 are amended. Claims 16-20 are added without the addition of new matter. Claims 1-20 are pending.
Response to Arguments
Applicant's arguments filed 08/14/2026 have been fully considered but they are not persuasive.
Applicant has argued that Bunandar fails to teach a respective analogue DC offset generator arranged to generate a respective DC offset signal encoding a constant DC term for the respective light detector. The examiner respectfully disagrees.
Claim 5 recites “a respective offset light generator configured to generate a respective offset optical signal”, mapped to Bunandar light source 202 that passes light to amplitude modulators 204, and “wherein the DC offset signal is generated by detecting the intensity of the offset optical signal by the offset photodetector”, mapped to optical detectors 210 and the photocurrent produced thereby. The amendment to claim 5 recites that the claimed analogue DC offset generator is comprised of the mapped elements, therefore Bunandar discloses the DC offset generator.
No other substantial arguments were presented after page 9 of Remarks. Therefore the prior art teaches the invention as currently claimed.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1-5, 8-18, and 20 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Bunandar et al. US PGPub 2021/036783 A1 (of record, see IDS dated 08/16/2023, hereinafter, “Bunandar”).
Regarding independent claim 1, Bunandar discloses a system for performing vector multiplication using optics (refer to title and abstract disclosing systems and methods for performing signed matrix operations using a linear photonic processor, and Fig. 1 shows photonic processing system 100 with photonic processor 110, par. [0037]), the system comprising one or more channels (Figs. 2 and 3A show a plurality of channels, with one channel for each vector), each comprising:
a respective light signal generator arranged to generate a respective optical signal (Fig. 2, linear photonic processor 200, which may be implemented as the optical source 108 and photonic processor 110 of photonic processing system 100 of Fig. 1, has light sources 202 which produce light passed to amplitude modulators 204, pars. [0045-46]);
a respective optical vector multiplier arranged to receive a vector of optical signals including the respective optical signal (Figs. 1 and 2, photonic processor 110 performs matrix-vector, matrix-matrix, and/or tensor-tensor multiplication operations with amplitude modulators 204 and 208 that encode elements of the input vector, matrix, and/or tensor, par. [0042]), and multiply by a respective vector of weights in an optical domain (Fig. 2, modulators 208 produce an optical signal having an intensity that represents a multiplication of the matrix element wpj and vector element xj in an amplitude of the output optical signal, par. [0050]), each optical signal having a modulated amplitude modelling a value of a respective variable from a vector of variables (Fig. 2, the modulation in amplitude to represent the value of a variable is performed by the amplitude modulators 204, par. [0046]), and the weights modelling interactions between the variables (amplitude modulators 208 produce a second optical signal with an optical intensity that represents a multiplication of the matrix element wpj and vector element in xj as an amplitude of the output optical signal, par. [0050], equivalent to a weights model for the interactions of variables); and
a respective light detector arranged to detect an intensity of a resulting output of the respective optical vector multiplier by incoherent detection (output second optical signal that is output by each second amplitude modulator 208 may be transmitted to and detected using an optical detector 210 which are photodetectors, par. [0051], and Bunandar teaches photonic processing architecture can utilize incoherent light, e.g., light in which the electromagnetic waves do not maintain a fixed and predictable phase relationship with each other over a period of time for its operations, par. [0035], and Fig. 1, optical source 108 may include, for example, one or more coherent and/or incoherent light sources configured to produce the N optical signals, par. [0040], and Fig. 2, light sources 202 may produce coherent or incoherent light that is passed to Q first amplitude modulators 204, par. [0046] and Bunandar teaches that to prevent unwanted interference between the signals routed to the same detector, a sufficiently incoherent light source can be used, par. [0053]), thereby generating an analogue intensity signal modulated on a scale that can only take positive values (optical detector 210 produce a photocurrent that is proportional to the intensity of light incident on the detector, par. [0051], and Bunandar teaches intensity is a non-negative number and discloses a method for performing a matrix-vector multiplication operation between a real-valued matrix and a real-valued vector using only amplitude-modulation by offsetting and/or rescaling the number line, par. [0094]);
a respective analogue DC offset generator arranged to generate a respective DC offset signal encoding a constant DC term for the respective light detector (Fig. 2, light source 202 produces light passed to amplitude modulators 204, par. [0046], and optical detectors 210 are photodetectors, par. [0051], that produce a photocurrent that is proportional to the intensity of light incident on the detector, par. [0051], and output vector yi can be computed using matrix M’ and vector x’ as described in par. [0096], where output vector yi has terms cMSjxj and cxSjMij that may be evaluated by a digital vector processor and Bunandar teaches these terms are subtracted to arrive at output vector yi, pars. [0093-98], thus Bunandar discloses an analogue DC offset generator); and
a respective differentiator arranged to subtract the respective DC offset signal from the analogue intensity signal, in order to produce a respective output signal in the form of an analogue electronic signal modulated on a scale having positive and negative values (output vector yi can be computed using matrix M’ and vector x’ as described in par. [0096], where output vector yi has terms cMSjxj and cxSjMij that may be evaluated by a digital vector processor and Bunandar teaches these terms are subtracted to arrive at output vector yi, pars. [0093-98]).
Regarding dependent claim 2, Bunandar discloses the system of claim 1, wherein the variables are binary (photonic processing system 100 receives, as an input from an external processor, e.g., a CPU, an input vector and/or matrix represented by a group of input bit strings and produces an output vector and/or matrix represented by a group of output bit strings, par. [0037], where a bit string is a sequence of binary digits).
Regarding dependent claim 3, Bunandar discloses the system of claim 1, wherein the optical vector multiplier (Figs. 1 and 2, photonic processor 110 performs matrix-vector, matrix-matrix, and/or tensor-tensor multiplication operations with amplitude modulators 204 and 208 that encode elements of the input vector, matrix, and/or tensor, par. [0042]) in each channel comprises one of:
- a spatial light modulator,
- a wavelength selective switch (Fig. 5, linear photonic processor 500, which is similar to linear photonic processors illustrated in Figs. 2, 3A, and 3B, includes wavelength division multiplexing circuit 512, pars. [0108-110], equivalent to a wavelength selective switch),
- a ring resonator (Figs. 4A-4C show illustrative examples of amplitude modulators that may be used in some embodiments as first and/or second amplitude modulators as described in connection with the examples of Figs. 2, 3A, and 3B, and in Fig. 4C shows a modulator 400c that uses a ring resonator 414, par. [0100]), or
- a Mach-Zehnder interferometer.
Regarding dependent claim 4, Bunandar discloses the system of claim 1, wherein the optical vector multiplier in at least one channel comprises a wavelength selective switch (Fig. 5, linear photonic processor 500, which is similar to linear photonic processors illustrated in Figs. 2, 3A, and 3B, includes wavelength division multiplexing circuit 512, pars. [0108-110], equivalent to a wavelength selective switch).
Regarding amended dependent claim 5, Bunandar discloses the system of claim 1, wherein the respective analogue DC offset generator comprises:
a respective offset light generator configured to generate a respective offset optical signal (Fig. 2, light source 202 produces light passed to amplitude modulators 204, par. [0046]), and a respective offset photodetector (optical detectors 210 are photodetectors, par. [0051]), arranged to detect the intensity of the offset optical signal, thereby generating the DC offset signal (the output second optical signal that is output by each second amplitude modulator 208 may be transmitted to and detected using an optical detector 210, par. [0051]), wherein the DC offset signal is generated by detecting the intensity of the offset optical signal by the offset photodetector (optical detectors 210 produce a photocurrent that is proportional to the intensity of light incident on the detector, par. [0051], and output vector yi can be computed using matrix M’ and vector x’ as described in par. [0096], where output vector yi has terms cMSjxj and cxSjMij that may be evaluated by a digital vector processor and Bunandar teaches these terms are subtracted to arrive at output vector yi, pars. [0093-98]).
Regarding amended dependent claim 8, Bunandar discloses the system of claim 1, wherein each channel comprises a respective feedback path arranged to return a respective feedback signal based on the respective output signal to the respective light signal generator, wherein the respective light signal generator is arranged to adapt the respective optical signal in dependence on the feedback signal (Bunandar teaches the linear photonic processor disclosed therein can accelerate various general matrix-matrix operations, par. [0085], and is amenable for back-propagation, i.e., training, of an artificial neural network, pars. [0087-90], where Examiner understands training to be functionally equivalent to a feedback path).
Regarding dependent claim 9, Bunandar discloses the system of claim 8, wherein in each channel the respective feedback path is arranged to add a respective noise component to the respective output signal in order to produce the respective feedback signal before return to the respective light signal generator (Bunandar teaches that in back-propagation with conventional stochastic gradient descent, the weight matrix is adjusted iteratively such that the weight matrix at time t + l is defined as a function of the weight matrix at time t and a derivative of the loss function with respect to the weights of the weight matrix, par. [0087], equivalent to a feedback path with a noise component added to an output signal).
Regarding dependent claim 10, Bunandar discloses the system of claim 9, the system arranged to estimate values of the vector of variables that optimize a function, the function comprising a weighted sum of a plurality of terms, each term comprising a product of a corresponding subset of the variables from said vector and each term being weighted by a corresponding weight from a matrix of weights that models interactions between the variables; wherein the respective vector of weights in each channel comprises a respective vector of weights from the matrix of weights, representing an interaction between the respective variable and the vector of variables (Bunandar teaches training an artificial neural network typically involves running a back-propagation algorithm, par. [0087], and discusses the relationship between an input vector x, a weight matrix w, and the updates Dw of the gradient of the loss function E in pars. [0085-91], equivalent to the limitation as currently written).
Regarding dependent claim 11, Bunandar discloses the system of claim 10, wherein the respective light signal generator in each channel is configured to perform the adaptation iteratively (Bunandar teaches training an artificial neural network typically involves running a back-propagation algorithm, par. [0087], and discusses the relationship between an input vector x, a weight matrix w, and the updates Dw of the gradient of the loss function E in pars. [0085-91],)
Bunandar does not explicitly or specifically disclose the adaption is iteratively performed according to:
x
i
k
+
1
=
cos
2
(
α
*
x
i
k
+
β
*
∑
j
J
i
j
*
x
j
k
-
π
4
+
ζ
i
[
k
]
)
-
1
2
where xi is the spin signal of channel i, k is in index of the iteration, a and b are coefficients, J is the matrix of weights, and z is the noise component.
A claim that recites a numerical formula or equation will be considered as falling within the "mathematical concepts" grouping, per MPEP 2106.04(a)(2)(B). Further, a claim that recites a mathematical calculation, when the claim is given its broadest reasonable interpretation in light of the specification, will be considered as falling within the "mathematical concepts" grouping. A mathematical calculation is a mathematical operation (such as multiplication) or an act of calculating using mathematical methods to determine a variable or number, e.g., performing an arithmetic operation such as exponentiation, refer to MPEP 2106.04(a)(2)(C).
As understood by the Examiner, the limitations reciting an iterative performance of a calculation directed to an adaptation is directed to organizing information and manipulating information through mathematical correlations. Per MPEP 2106.04(a)(2)(I)(A) (i), examples of mathematical relationships recited in a claim using an algorithm for determining the optimal number of visits by a business representative to a client, In re Maucorps, 609 F.2d 481, 482, 203 USPQ 812, 813 (CCPA 1979). In this case, the formula for iteration is, as best understood by the Examiner, directed to an optimization process for calculations, and Bunandar teaches iterative calculations of parameters in the context of a linear photonic processor. Therefore, a person of ordinary skill, before the effective filing date of the claimed invention, would be aware of the method of performing iterative calculations for optimizing a parameter in a linear photonic processor, and would be able to determine an algorithm, formula, or equation to arrive at an optimal value.
Regarding dependent claim 12, Bunandar discloses the system of claim 10, the system comprising a plurality of said channels, wherein:
the amplitude of the respective optical signal generated by the respective light signal generator in each channel is modulated to model the value of different respective one of the variable from said vector of variables (Fig. 2, linear photonic processor 200, the modulation in amplitude of signals from light sources 202 to represent the value of a variable is performed by the amplitude modulators 204, par. [0045-46}); and
each channel further comprises a respective splitter arranged to supply an instance of the respective optical signal to each of the plurality of channels, the optical vector multiplier in each channel thus receiving the vector of optical signals in order to perform the respective vector multiplication (Fig. 3A, linear photonic processor 300a, which is similar to linear photonic processor 200, see Fig. 2, includes light source 302 that outputs an optical signal that passes through beam splitter 304, par. [0063]).
Regarding amended dependent claim 13, Bunandar discloses the system of claim 10, the system comprising a single channel in which the light signal generator is configured to multiplex the optical signals into a same beam of light by time-division multiplexing; wherein the optical vector multiplier comprises an arrangement of delay lines to delay the optical signals of said vector by different path lengths so as to overlap in time, and at least one further optical element arranged to perform the vector multiplication based on the delayed optical signals (Bunandar teaches the option of time-multiplexing the optical detectors 210 such that a first output signal arrives first, a second output signal arrives after the first, a third output signal arrives after the second, and so on, par. [0053]).
Regarding dependent claim 14, Bunandar discloses the system of claim 1, wherein each channel is used to represent a node or layer of a neural network (Bunandar teaches the linear photonic processor disclosed therein is applicable to artificial neural networks with multiple layers of tensor-tensor multiplications, pars. [0085-91]).
Regarding amended independent claim 15, Bunandar discloses a method of performing vector multiplication using optics (refer to title and abstract disclosing systems and methods for performing signed matrix operations using a linear photonic processor, and Fig. 1 shows photonic processing system 100 with photonic processor 110, par. [0037]), the method comprising, for each of one or more channels (Figs. 2 and 3A show a plurality of channels for each vector):
generating a respective optical signal (Fig. 2, light source 202 produces light passed to amplitude modulators 204, par. [0046]);
receiving, at a respective optical vector multiplier, a vector of optical signals, including the respective optical signal (Figs. 1 and 2, photonic processor 110 performs matrix-vector, matrix-matrix, and/or tensor-tensor multiplication operations with amplitude modulators 204 and 208 that encode elements of the input vector, matrix, and/or tensor, par. [0042]),
multiplying the vector of optical signals by a respective vector of weights in an Fig. 2, modulators 208 produce an optical signal having an intensity that represents a multiplication of the matrix element wpj and vector element xj in an amplitude of the output optical signal, par. [0050]), each optical signal having a modulated amplitude modelling a value of a respective variable from a vector of variables (Fig. 2, the modulation in amplitude to represent the value of a variable is performed by the amplitude modulators 204 at least, par. [0046]), and the weights modelling interactions between the variables (optical intensity represents a multiplication of the matrix element wpj and vector element in xj an amplitude of the output optical signal, par. [0050], equivalent to a weights model for the interactions of variables);
detecting, at a respective light detector, an intensity of a resulting output of the respective optical vector multiplier by incoherent detection (a photonic processing architecture can utilize incoherent light, e.g., light in which the electromagnetic waves do not maintain a fixed and predictable phase relationship with each other over a period of time for its operations, par. [0035], and optical source 108 may include, for example, one or more coherent and/or incoherent light sources configured to produce the N optical signals, par. [0040], and light sources 202 may produce coherent or incoherent light that is passed to Q first amplitude modulators 204, par. [0046] and Bunandar teaches that to prevent unwanted interference between the signals routed to the same detector, a sufficiently incoherent light source can be used, par. [0053]), thereby generating an analogue intensity signal modulated on a scale that can only take positive values (Bunandar teaches intensity is a non-negative number, and therefore discloses a method for performing a matrix-vector multiplication operation between a real-valued matrix and a real-valued vector using only amplitude-modulation by offsetting and/or rescaling the number line, par. [0094]);
generating, with a respective analogue DC offset generator, a respective DC offset signal encoding a constant DC term for the respective light detector (Fig. 2, light source 202 produces light passed to amplitude modulators 204, par. [0046], and optical detectors 210 are photodetectors, par. [0051], that produce a photocurrent that is proportional to the intensity of light incident on the detector, par. [0051], and output vector yi can be computed using matrix M’ and vector x’ as described in par. [0096], where output vector yi has terms cMSjxj and cxSjMij that may be evaluated by a digital vector processor and Bunandar teaches these terms are subtracted to arrive at output vector yi, pars. [0093-98], thus Bunandar discloses an analogue DC offset generator); and
subtracting, at a respective differentiator, the respective DC offset signal from the analogue intensity signal, in order to produce a respective output signal in the form of an analogue electronic signal modulated on a scale having positive and negative values (output vector yi can be computed using matrix M’ and vector x’ as described in par. [0096], where output vector yi has terms cMSjxj and cxSjMij that may be evaluated by a digital vector processor and Bunandar teaches these terms are subtracted to arrive at output vector yi, pars. [0093-98]).
Claims 6-7 are rejected under 35 U.S.C. 103 as being unpatentable over Bunandar as applied to claim 1 above, in view of Koyama et al US PGPub 2020/0074279 A1 (of record, see Office action dated 05/28/2026, hereinafter, “Koyama”).
Regarding amended dependent claim 6, Bunandar discloses the system of claim 1, wherein in each channel:
the respective light signal generator (Fig. 2, linear photonic processor 200 has light sources 202, pars. [0045-46]) comprises a respective generator arranged to generate a respective signal in the form of an analogue electronic signal representing the respective variable (the photocurrent generated by optical detectors 210 combine to produce a summed output representing a product of a vector x and a row of matrix w, and the value of the summed photocurrent may be read out using a combination of a transimpedance amplifier and an analog-to-digital converter, see par. [0052]), and a modulator arranged to modulate the amplitude of the optical signal based on the respective analogue signal (Fig. 2, light sources 202 produce light passed to amplitude modulators 204, pars. [0045-46]); and
the respective signal varies on a scale between positive and negative levels to represent the respective variable (for most applications, the photonic processor will be performing a multiplication between a matrix and a vector that may include positive and/or negative-valued elements, par. [0094]), but the amplitude of the optical signal can only be positive (intensity is a non-negative number, par. [0094]), the modulator being configured to convert the positive and negative levels of the signal into positive amplitudes of the optical signal (Bunandar discloses a method for performing a matrix-vector multiplication operation between a real valued matrix and a real-valued vector using only amplitude-modulation by offsetting and/or rescaling the number line, par. [0094]).
Bunandar does not disclose a light signal generator that comprises a spin generator, and therefore does not disclose a spin signal.
In a related field of invention, Koyama discloses an optimization apparatus 10, shown in at least Fig. 1 thereof, where the optimization apparatus performs a calculation by replacing a problem to be calculated with an Ising model indicative of the behavior of the spin of a magnetic material (par. [0003] thereof), and apparatus 10 includes a plurality of Ising units 11a which perform calculations regarding neurons (par. [0041] thereof), where the state of each neuron is represented by the use of a variable having a plurality of values (par. [0032] thereof).
Therefore, it would have been obvious to a person having ordinary skill in the art, before the effective filing date of the claimed invention, to have applied the teachings of Koyama to the disclosure of Bunandar and included a spin generator in the light sources 202 of the photonic processing circuit, because Koyama teaches that by doing so, the structure of circuits used for calculating an energy change of the Ising model is simplified and an increase in circuit scale is suppressed (Koyama, par. [0075]).
Regarding dependent claim 7, Bunandar in view of Koyama discloses the system of claim 6, but the prior art combination does not specifically disclose wherein the respective spin generator in each channel comprises a further light source, a further modulator arranged to modulate light from the further light source in dependence on a respective feedback signal, and a further light detector arranged to detect the modulated light from the further modulator and generate the spin signal in dependence thereon.
The prior art combination discloses the claimed invention except for further light sources (Bunandar discloses a plurality of light sources 202, pars. [0045-46]), modulators (Bunandar discloses a plurality of amplitude modulators 204, pars. [0045-46]), and light detectors (Bunandar discloses a plurality of optical detectors 210, par. [0051]). Therefore, it would have been obvious to one of ordinary skill in the art at the time the invention was made to include further light sources, modulators, and detectors, since it has been held that mere duplication of the essential working parts of a device involves only routine skill in the art. St. Regis Paper Co. v. Bemis Co., 193 USPQ 8 (1977). In this case, the prior art reference Bunandar discloses a plurality of light sources, modulators, and light detectors, and it would be obvious to, and within the ordinary skill of, a person having ordinary skill to include more elements that are already disclosed in the device taught by Bunandar.
Regarding new dependent claim 16, Bunandar discloses the system of claim 1, wherein the respective light signal generator (Fig. 2, linear photonic processor 200, which may be implemented as the optical source 108 and photonic processor 110 of photonic processing system 100 of Fig. 1, has light sources 202 which produce light passed to amplitude modulators 204, pars. [0045-46]) comprises:
a light source arranged to generate the respective optical signal (Fig. 2, light sources 202, par. [0045-46]); and a
modulator arranged to modulate the amplitude of the respective optical signal to model the value of the respective variable (Fig. 2, amplitude modulators 204, pars. [0045-46]).
Regarding new independent claim 17, Bunandar discloses an optical optimization system comprising:
a plurality of channels (Figs. 2 and 3A show a plurality of channels, with one channel for each vector); and
a plurality of feedback loops (Bunandar teaches the linear photonic processor disclosed therein can accelerate various general matrix-matrix operations, par. [0085], and is amenable for back-propagation, i.e., training, of an artificial neural network, pars. [0087-90], where Examiner understands training to be functionally equivalent to a feedback path);
wherein each channel comprises:
a light signal generator arranged to generate an optical signal having an amplitude modelling a value of a variable from a vector of variables (Fig. 2, linear photonic processor 200, which may be implemented as the optical source 108 and photonic processor 110 of photonic processing system 100 of Fig. 1, has light sources 202 which produce light passed to amplitude modulators 204, pars. [0045-46]);
an optical vector multiplier arranged to receive a vector of optical signals including the optical signal, and multiply the vector of optical signals by a vector of weights in an optical domain, the weights modelling interactions between the variables of the vector of variables (Figs. 1 and 2, photonic processor 110 performs matrix-vector, matrix-matrix, and/or tensor-tensor multiplication operations with amplitude modulators 204 and 208 that encode elements of the input vector, matrix, and/or tensor, par. [0042]);
a light detector arranged to detect an intensity of a resulting output of the optical vector multiplier by incoherent detection, thereby generating an analogue intensity signal modulated on a scale that can only take positive values (optical signal that is output by each second amplitude modulator 208 may be transmitted to and detected using an optical detector 210 which are photodetectors, par. [0051], and Bunandar teaches photonic processing architecture can utilize incoherent light, e.g., light in which the electromagnetic waves do not maintain a fixed and predictable phase relationship with each other over a period of time for its operations, par. [0035], and Fig. 1, optical source 108 may include, for example, one or more coherent and/or incoherent light sources configured to produce the N optical signals, par. [0040], and Fig. 2, light sources 202 may produce coherent or incoherent light that is passed to Q first amplitude modulators 204, par. [0046] and Bunandar teaches that to prevent unwanted interference between the signals routed to the same detector, a sufficiently incoherent light source can be used, par. [0053]);
an analog DC offset generator arranged to generate a DC offset signal encoding a constant DC term for the light detector (Fig. 2, light source 202 produces light passed to amplitude modulators 204, par. [0046], and optical detectors 210 are photodetectors, par. [0051], that produce a photocurrent that is proportional to the intensity of light incident on the detector, par. [0051], and output vector yi can be computed using matrix M’ and vector x’ as described in par. [0096], where output vector yi has terms cMSjxj and cxSjMij that may be evaluated by a digital vector processor and Bunandar teaches these terms are subtracted to arrive at output vector yi, pars. [0093-98], thus Bunandar discloses an analogue DC offset generator);
a differentiator arranged to subtract the DC offset signal from the analogue intensity signal, thereby producing an output signal in the form of an analogue electronic signal modulated on a scale having positive and negative values (output vector yi can be computed using matrix M’ and vector x’ as described in par. [0096], where output vector yi has terms cMSjxj and cxSjMij that may be evaluated by a digital vector processor and Bunandar teaches these terms are subtracted to arrive at output vector yi, pars. [0093-98]); and
a feedback path arranged to return a feedback signal based on the output signal to the light signal generator, the light signal generator arranged to adapt the optical signal in dependence on the feedback signal (Bunandar teaches the linear photonic processor disclosed therein can accelerate various general matrix-matrix operations, par. [0085], and is amenable for back-propagation, i.e., training, of an artificial neural network, pars. [0087-90], where Examiner understands training to be functionally equivalent to a feedback path);
wherein the feedback path of each channel forms one of the plurality of feedback loops, and the plurality of feedback loops are arranged to iteratively adapt the optical signals of the plurality of channels in the optical and analogue electronic domains (Bunandar teaches the linear photonic processor disclosed therein can accelerate various general matrix-matrix operations, par. [0085], and is amenable for back-propagation, i.e., training, of an artificial neural network, pars. [0087-90], where Examiner understands training to be functionally equivalent to a feedback path).
Regarding new dependent claim 18, Bunandar discloses the optical optimization system of claim 17, wherein the light signal generator comprises:
a light source arranged to generate the optical signal ( Fig. 2, linear photonic processor 200, which may be implemented as the optical source 108 and photonic processor 110 of photonic processing system 100 of Fig. 1, has light sources 202 which produce light passed to amplitude modulators 204, pars. [0045-46] ); and
a modulator arranged to modulate the amplitude of the optical signal to model the value of the variable (amplitude modulators 204, pars. [0045-46]).
Regarding new dependent claim 20, Bunandar discloses the optical optimization system of claim 17, wherein in each channel the feedback path is arranged to add a noise component to the output signal in order to produce the feedback signal before return to the light signal generator (Bunandar teaches that in back-propagation with conventional stochastic gradient descent, the weight matrix is adjusted iteratively such that the weight matrix at time t + l is defined as a function of the weight matrix at time t and a derivative of the loss function with respect to the weights of the weight matrix, par. [0087], equivalent to a feedback path with a noise component added to an output signal).
Allowable Subject Matter
Claim 19 is objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Regarding new dependent claim 19, Bunandar discloses the optical optimization system of claim 17, wherein the plurality of feedback loops are arranged to iteratively adapt the optical signals (Bunandar teaches the linear photonic processor disclosed therein can accelerate various general matrix-matrix operations, par. [0085], and is amenable for back-propagation, i.e., training, of an artificial neural network, pars. [0087-90], where Examiner understands training to be functionally equivalent to a feedback path).
Bunandar does not disclose the output signals of the plurality of channels settle to a stable state representing an assignment of values to the vector of variables that optimizes a function, the function comprising a weighted sum of a plurality of terms, each term comprising a product of a corresponding subset of the variables from said vector and each term being weighted by a corresponding weight from a matrix of weights that models the interactions between the variables.
Conclusion
Applicant's amendment necessitated the new grounds of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Justin W Hustoft whose telephone number is (571)272-4519. The examiner can normally be reached Monday - Friday 9:00 AM - 5:00 PM Eastern Time.
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, Ricky L Mack can be reached at (571)272-2333. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/JUSTIN W. HUSTOFT/Examiner, Art Unit 2872
/MARIN PICHLER/Primary Examiner, Art Unit 2872
