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 .
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 1-4 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
The examiner notes that the claims seems to be directed to a method for using random numbers as inputs for one or more supervised or unsupervised learning training examples for a neural network. Throughout the specification, the invention is described as introducing random data into the training process, which does not appear to be a concrete invention but rather, an aspirational concept. There is no disclosure of, for example, a specific training algorithm, a concrete loss function adaptation, a convergence mechanism, safeguards against training collapse, or any principled justification for why the method works. Written description requires more than stating what you want to achieve. The applicant must also show they possessed a particular solution. This appears to simply be directed to the notion of adding randomness.
Further, “Supervised Learning” appears to be contradicted rather than supported. The disclosure itself describes supervised learning at [0006] to require input to be paired with a desired output. However, the specification repeatedly states that outputs are random, randomly selected from the model’s own outputs, or randomly adjusted during training. Thus, the applicant does not demonstrate possession of a supervised learning method, but rather describes non-supervised or self-referential behavior without reconciling the contradiction. There is no explanation of how these desired outputs remain desired, how correctness is defined, or how learning signals remain meaningful.
Further, key elements are claimed functionally without structural or procedural support including “random data”, “random but valid input values”, “random but valid expected output values”, and “probabilities based on assigned values”. The specification never defines what “valid” means in a general neural network, how probability weights are assigned, how randomness interacts with loss minimization, or how training stability is maintained. It appears that a result (training with randomness) is being claimed without teaching the means.
Further, claims 1-4 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the enablement requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to enable one skilled in the art to which it pertains, or with which it is most nearly connected, to make and/or use the invention. The claims seem to cover any neural network, any supervised or unsupervised learning task, initial, periodic, or continuous retraining, any random input, and any random output generation method.
The disclosure, however, only provides basic numeric examples, no generalizable training guidance, and no boundaries on when the method works versus fails. A person of ordinary skill in the art would have to determine, through extensive experimentation, for example, how much random variance is tolerable, when randomness destabilizes gradients, how often retraining occurs, how accuracy is preserved, and which architectures are compatible. This is undue experimentation and not routine optimization.
Further, the specification never demonstrates that invention works in any real-world scenario. There is no empirical result, no convergence discussion, no example beyond trivial numeric ranges, and no explanation of why training wouldn’t degrade. If most embodiments would fail or degrade training, the claim is seemingly not enabled across its full scope.
Further, desired output selected by weighted probability selection among outputs generated during training is essentially training on the model’s own predictions with stochastic label assignment with no guidance provided in the specification on preventing feedback loops, avoiding mode collapse, maintaining learning signal integrity, or ensuring convergence. Therefore, a person of ordinary skill in the art would have no reasonable expectation of success practicing this.
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 1-4 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.
Claim 1 recites the following limitations: “the use of random data” “the input data”, “the expected output”, & “the assigned values” in “A method comprising the use of random data as one or more supervised learning training examples for a neural network…where the input data is random and the expected output is random or established by randomly selecting among model outputs with probabilities based on the assigned values of those outputs.”. There is insufficient antecedent basis for these limitations in the claim.
Claim 2 recites the following limitations: “the use of random data”, “the input values” & “the expected output” in “A method comprising the use of random data in one or more supervised learning training examples for a neural network…where the random data is used to modify the input values, the expected output, or both in an existing training example.”. There is insufficient antecedent basis for these limitations in the claim.
Claim 3 recites the following limitation: “the random adjustment of environment state transition probabilities or expected rewards” in “A method comprising the random adjustment of environment state transition probabilities or expected rewards...”. There is insufficient antecedent basis for this limitation in the claim.
Claim 4 recites the following limitations: “the introduction of random data” “the data set” in “A method comprising the introduction of random data into the data set used for initial or ongoing unsupervised training of a neural network...”. There is insufficient antecedent basis for these limitations in the claim.
Claims 1-4 are rejected as failing to define the invention in the manner required by 35 U.S.C. 112(b) or pre-AIA 35 U.S.C. 112, second paragraph.
The claims are narrative in form and replete with indefinite language. The structure which goes to make up the device must be clearly and positively specified. The structure must be organized and correlated in such a manner as to present a complete operative device. The claims must be in one sentence form only. Note the format of the claims in the patent(s) cited.
EXAMINER’S NOTE TO APPLICANT:
In the claims for an application, everything cited must be positively recited before it may be referenced. When citing something such as a data set, for example, by referring to “the data set” when “a data set” has not been previously positively recited within that claim, this is a basis for rejection under 35 U.S.C. 112(b). One could, for example, state that a method comprising a use of random data, further comprising input data, an expected output, and assigned values of those outputs, wherein… etc. The examiner would like to point the applicant to the MPEP 2173.05(e) for full details of the basis of these rejections.
Claim Rejections - 35 USC § 102
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 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.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1-2 are rejected under 35 U.S.C. 102(a)(1) as being clearly anticipated by Zhang, C. et al. “Understanding deep learning requires rethinking generalization.” (hereafter, ZHANG).
Regarding claim 1, ZHANG teaches “A method comprising the use of random data as one or more supervised learning training examples for a neural network either for initial training or periodic or continuous retraining…”:
([Introduction, 1.1 Randomization tests] “At the heart of our methodology is a variant of the well-known randomization test from non-parametric statistics (Edgington & Onghena, 2007). In a first set of experiments, we train several standard architectures on a copy of the data where the true labels were replaced by random labels.”) A standard architecture with true labels indicates supervised learning training examples for a neural network. They are explicitly shown to comprise random labels/data here. It is inherently for either initial training, or periodic/continuous retraining, as there are no alternative possibilities.
Further, ZHANG teaches “…where the input data is random”:
([2.1 Fitting Random Labels and Pixels] “We run our experiments with the following modifications of the labels and input images:
…
• Random pixels: a different random permutation is applied to each image independently.
• Gaussian: A Gaussian distribution (with matching mean and variance to the original image dataset) is used to generate random pixels for each image.”)
Further, ZHANG teaches “the expected output is random or established by randomly selecting among model outputs with probabilities based on the assigned values of those outputs”
([2.1 Fitting Random Labels and Pixels] “We run our experiments with the following modifications of the labels and input images:
• True labels: the original dataset without modification.
• Random labels: all the labels are replaced with random ones. (random expected outputs.)
Regarding claim 2, ZHANG teaches “A method comprising the use of random data in one or more supervised learning training examples for a neural network either for initial training or periodic or continuous retraining…”:
([Introduction, 1.1 Randomization tests] “At the heart of our methodology is a variant of the well-known randomization test from non-parametric statistics (Edgington & Onghena, 2007). In a first set of experiments, we train several standard architectures on a copy of the data where the true labels were replaced by random labels.”) A standard architecture with true labels indicates supervised learning training examples for a neural network. They are explicitly shown to comprise random labels/data here. It is inherently for either initial training, or periodic/continuous retraining, as there are no alternative possibilities.
Further, ZHANG teaches “…where the random data is used to modify the input values, the expected output, or both in an existing training example”:
([2.1 Fitting Random Labels and Pixels] “We run our experiments with the following modifications of the labels and input images:
…
• Shuffled pixels: a random permutation of the pixels (random data) is chosen and then the same permutation is applied to all the images in both training and test set (the random data is used to modify the input values and the expected output).”)
Claim 3 is rejected under 35 U.S.C. 102(a)(1) as being clearly anticipated by Wang, J. et al. “Reinforcement Learning with Perturbed Rewards.” (hereafter, WANG)
Regarding claim 3, WANG teaches “A method comprising the random adjustment of environment state transition probabilities or expected rewards during initial or ongoing reinforcement training of a neural network”:
([Abstract] “Recent studies have shown that reinforcement learning (RL) models are vulnerable in various noisy scenarios… In this paper, we consider noisy RL problems with perturbed rewards (randomly adjusted expected rewards during reinforcement training of a neural network), which can be approximated with a confusion matrix.”)
Claim 4 is rejected under 35 U.S.C. 102(a)(1) as being clearly anticipated by Vincent, P. et al. “Extracting and Composing Robust Features with Denoising Autoencoders.” (hereafter, VINCENT)
Regarding claim 4, VINCENT teaches “A method comprising the introduction of random data into the data set used for initial or ongoing unsupervised training of a neural network either as a separate random example or as a random variation applied to an existing example”:
([Abstract] “difficulties in learning deep generative or discriminative models can be overcome by an initial unsupervised learning step that maps inputs to useful intermediate representations. We introduce and motivate a new training principle for unsupervised learning of a representation based on the idea of making the learned representations robust to partial corruption of the input pattern… Comparative experiments clearly show the surprising advantage of corrupting the input of autoencoders on a pattern classification benchmark suite.”)
And further detailed at:
([Page 1098, 2.3 The Denoising Autoencoder] “We will now train it (the autoencoder) to reconstruct a clean “repaired” input from a corrupted, partially destroyed one. This is done by first corrupting the initial input… to get a partially destroyed version… by means of a stochastic mapping… In our experiments, we considered the following corrupting process, parameterized by the desired proportion... of “destruction”: for each input…, a fixed number… of components are chosen at random, and their value is forced to 0, while the others are left untouched.”) Here, we see that random data is introduced as a random variation applied to an existing example during unsupervised training.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MATTHEW LEE LEWIS whose telephone number is (571)272-1906. The examiner can normally be reached Monday: 12:00PM - 4:00PM and Tuesday - Friday: 12:00PM - 9PM.
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, Tamara Kyle can be reached at (571)272-4241. 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.
/TAMARA T KYLE/Supervisory Patent Examiner, Art Unit 2144
/Matthew Lee Lewis/Examiner, Art Unit 2144