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 § 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.
Claim(s) 1, 3, 7, 8, 13, and 14 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Kamada, Shin, and Takumi Ichimura. "An adaptive learning method of deep belief network by layer generation algorithm."
Regarding claim 1, Kamda teaches “a computer-implemented method comprising: training a first deep belief network (DBN) using training data” (pg. 2 ¶1 “The RBM trains the weights and some parameters for visible and hidden neurons to reach the small value of energy function.”);
“adding at least one neuron to the first DBN to generate a second DBN” (pg. 2 §B ¶1 “The basic idea of the neuron generation and neuron annihilation algorithm works in multi layered neural network [11] and we introduce the concept into RBM training method” and ¶2 “a new hidden neuron is added to the RBM to split the fluctuated neuron by inheritance of the parent hidden neuron attributes”);
“assigning weights to neurons in the second DBN according to weights of neurons in the trained first DBN” (pg. 2
PNG
media_image1.png
250
486
media_image1.png
Greyscale
the new neuron corresponds to the previous neuros and therefore the weight is determined based on the first weights and pg. 2 §B ¶2 “a new hidden neuron is added to the RBM to split the fluctuated neuron by inheritance of the parent hidden neuron attributes”); and
“training the second DBN using the training data” (previous citation §B ¶1 “The basic idea of the neuron generation and neuron annihilation algorithm works in multi layered neural network [11] and we introduce the concept into RBM training method”).
Regarding claim 3, Kamada teaches “wherein the assigning of the weights to the neurons in the second DBN according to the weights of the neurons in the trained first DBN comprises: for the at least one neuron that is newly added in the second DBN, assigning a weight of a neuron sampled from the trained first DBN” (pg. 2 §B ¶2 “a new hidden neuron is added to the RBM to split the fluctuated neuron by inheritance of the parent hidden neuron attributes”)
Regarding claim 7, Kamda teaches “further comprising: adding at least one neuron to the second DBN to generate a third DBN; assigning weights to neurons in the third DBN according to the weights of the neurons in the trained second DBN; and training the third DBN using the training data” (pg. 2 §B “The adaptive learning method of RBM [10] can be self organized structure to discover the optimal number of hidden neurons and weights according to the features of a given input data set in learning phase. The basic idea of the neuron generation and neuron annihilation algorithm works in multi layered neural network [11] and we introduce the concept into RBM training method.” the technique is not limited to a single neuron and is capable and functions the same over multiple neurons if needed)
Regarding claim 8, Kamda further teaches “further comprising: iterating the adding of the at least one neuron to the first DBN, the assigning of the weights to the neurons in the second DBN, and the training of the second DBN, wherein the first DBN of each iteration is the second DBN of a previous iteration” (pg. 2 §B “The adaptive learning method of RBM [10] can be self organized structure to discover the optimal number of hidden neurons and weights according to the features of a given input data set in learning phase. The basic idea of the neuron generation and neuron annihilation algorithm works in multi layered neural network [11] and we introduce the concept into RBM training method.” the technique is not limited to a single neuron and is capable and functions the same over multiple neurons if needed)
Independent claims 13 and 14 recite the same substantial subject matter as independent claim 1, only differing in embodiment. The differences in embodiment, a CRM and system compared to a method are obvious variations of another and therefore the claims are subject to the same rejection.
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.
Claim(s) 2, 4-6, 10, 11 is/are rejected under 35 U.S.C. 103 as being unpatentable over Kamada in view of Kurup, Aswathy Rajendra, Meenu Ajith, and Manel Martínez Ramón. "Semi-supervised facial expression recognition using reduced spatial features and deep belief networks."
Regarding claim 2, the Kamada reference has been addressed above. More specifically, Kurup teaches the limitations “wherein the training of the first DBN comprises training the first DBN for a predefined number of epochs or for a predefined amount of time” (Kurup pg. 5 §3.2 “The main test parameters used were the learning rate, the number of epochs, reconstruction error, and accuracy” epochs are predetermined as per hyperparameters, ubiquitous in machine learning)
It would have been obvious to one having ordinary skill in the art at the time that the invention was effectively filed to combine the teachings of Kamdaa with that of Kurup since a combination of known methods would yield predictable results. Kamada pertains to machine learning but doesn’t necessarily show the lower level details that Kurup specifically mentions. There concepts, epochs, learning rate, error, are standard machine learning parameters and have the same use in Kamada.
Regarding claim 4, the Kamada and Kurup references have been addressed above. Kurup further teaches “wherein the training of the second DBN comprises training the second DBN for a predefined number of epochs or for a predefined amount of time” (Kurup pg. 5 §3.2 “The main test parameters used were the learning rate, the number of epochs, reconstruction error, and accuracy” epochs are predetermined, this is not limited to any instance of the network)
Regarding claim 5, the Kamada and Kurup references have been addressed above. Kurup further teaches “further comprising: computing a reconstruction error after a number of training epochs when the training of the second DBN occurs or at a time interval during the training of the second DBN” (Kurup pg. 5 §3.2 “The main test parameters used were the learning rate, the number of epochs, reconstruction error, and accuracy)
Regarding claim 6, the Kamada and Kurup references have been addressed above. Kurup further teaches “further comprising: ending the training of the second DBN after an epoch provided an absolute difference between successive reconstruction errors is less than a threshold value” (Kurup pg. 4 fig. 3 “Architecture of the semi-supervised DBN. The structure consists of five layers. The number of nodes, from input to output layers, is 2, 3, 3, 4 nodes. The number of staked RBMs is 4. The dashed area highlights the structure of the first stacked RBMs. The network is first trained with unlabelled data using CD until convergence, and then a softmax activation of the output layer is used to apply a BP with the labelled data.”)
Regarding claim 10, the Kamada and Kurup references have been addressed above. Kurup further teaches “wherein the trained second DBN is useable for object detection or tracking in association with video data” (Kurup abstract “A semi-supervised emotion recognition algorithm using reduced features as well as a novel feature selection approach is proposed. The proposed algorithm consists of a cascaded structure where first a feature extraction is applied to the facial images, followed by a feature reduction.”)
Regarding claim 11, the Kamada and Kurup references have been addressed above. Kurup further teaches “wherein training the first DBN and the second DBN comprises training the first DBN and the second DBN to classify a wellbeing state of an individual user based on the training data according to a self-reported wellbeing state as ground truth data” (previous citation, emotion recognition or wellbeing and pg. 5 §5.1 “The extended Cohn-Kanade database (CK + ) [49] , the Radboud Faces database (RaFD) [50] and MMI database [51] were used to test the proposed method for facial ER. The CK + and MMI databases were captured from a lab-based environment whereas the RaFD database contained facial images with varying poses and gaze directions.”)
Claim(s) 9 is/are rejected under 35 U.S.C. 103 as being unpatentable over Kamada in view of Ma, Yuanjing, Junshuang Li, and Ruifeng Guo. "Application of data fusion based on deep belief network in air quality monitoring."
Regarding claim 9, the Kamada reference has been addressed above. Kamada does not explicitly teach the claim limitations. Ma however teaches “wherein the trained second DBN is to classify a wellbeing state of an individual user based on sensor data including any of: air pollution; particulate matter concentration; oxidised gas concentration; reduced gas concentration; NH3 concentration; noise level; people count in a vicinity; electrodermal activity; heart rate; heart rate variability; body temperature; blood volume pulse; and body movement” (Ma abstract “The successive development of social industrialization and modernization has caused a great many of environmental problems.Due to the imbalance between the environment and rapid development, the problem of urban air quality has become more and more prominent, so air quality monitoring has become particularly important. In air quality monitoring, unexpected situations such as damage to the monitoring equipment and the relocation of the station building will occur, resulting in the lack of monitoring data. For this lack of monitoring data, the missing values of the monitoring data are supplemented by the data fusion method using deep belief network (DBN). It can provide relevant researchers with reference data for further analysis and research on air quality. Compared with BP neural network for data fusion, DBN method is closer to the actual monitoring value.”)
It would have been obvious to one having ordinary skill in the art at the time that the invention was effectively filed to combine the teachings of Kamada with that of Ma since a combination of known methods would yield predictable results. As shown in Ma, DBNs are known to work well with air processing and monitoring. This is an established application of the networks and therefore would operate with the DBNs as discussed above.
Claim(s) 12 and 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Kamada in view of Wu, Jiajun, et al. "An energy-efficient deep belief network processor based on heterogeneous multi-core architecture with transposable memory and on-chip learning."
Regarding claims 12 and 15, the Kamada reference has been addressed above. While Kamada isn’t limited to any hardware, more specifically, Wu teaches “wherein the computer-implemented method is implemented on heterogeneous hardware with a plurality of processors operating in parallel” (Wu abstract “This paper proposes an energy-efficient processor design based on Deep Belief Network (DBN), which is one of the most suitable DNN models for on-chip learning”)
It would have been obvious to one having ordinary skill in the art at the time that the invention was effectively filed to combine the teachings of Kamada with that of Wu since it is shown that DBNs can be ran on a variety of hardware.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KEVIN W FIGUEROA whose telephone number is (571)272-4623. The examiner can normally be reached Monday-Friday, 10AM-6PM EST.
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, MIRANDA HUANG can be reached at (571)270-7092. 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.
KEVIN W FIGUEROA
Primary Examiner
Art Unit 2124
/Kevin W Figueroa/Primary Examiner, Art Unit 2124
Prosecution Insights