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 .
DETAILED ACTION
2. This office action is in response to the amendment filed on 07/21/2026. Claims 1-10 are canceled and claims 11-19 are pending and have been considered below.
3. The rejections of claim19 under 35 U.S.C. 112(a) or pre-AIA 35 U.S.C. 112, first paragraph are moot pursuant to amendments.
4. The rejections of claims 11-19 under 35 U.S.C. 101 as the claimed invention is directed to non-statutory subject matter are moot pursuant to amendments and arguments.
Claim Rejections - 35 USC § 103
5. 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.
Claims 11-19 are rejected under 35 U.S.C. 103 as being unpatentable over Houlsby et al. (US 2022/0092416) in view of Benyahia et al. (US 2020/0104688).
Claim 11. Houlsby discloses a computer-implemented method for creating a machine learning system, comprising the following steps:
providing a directed graph having an input node and output node connected by a plurality of edges and nodes (a graph whose nodes represent architectural decision points and whose edges represent available actions. Each architecture corresponds to a path from an initial node to a terminal node)([0034]–[0038]);
randomly drawing a plurality of paths through the directed graph along drawn edges of the directed graph (generating batches of paths using a controller whose current parameters determine the sampling distributions)( [0040], [0042]–[0043] and [0071])…( Its path-generation procedure repeatedly samples actions from those distributions)([0082]-[0086]), wherein each respective path of the plurality of paths is drawn by successively drawing edges, wherein at each node reached during the drawing of a respective path (Houlsby's controller generates a score distribution over outgoing edges, samples an action and adds the node reached by the sampled edge to the path… Its detailed path-generation procedure expressly permits the scores to be probabilities) ([0040], [0084]-[0086]);
training machine learning systems corresponding to the drawn plurality of paths, wherein parameters of the machine learning systems are adjusted during training so that a cost function is optimized (generating a task-network instance corresponding to each candidate architecture and trains it on task-specific training data)([0043]-[0044], [0071]-[0074])..( training the controller using performance-based rewards and describes maximizing expected reward through policy-gradient optimization) ([0078]–[0080]), wherein the parameters that are adjusted include the probabilities of the edges of the drawn plurality of paths (controller has trainable weights and biases that determine the score distributions used to sample outgoing edges. Its parameter-updating engine adjusts those parameters using evaluations of generated architectures)([0040],[0042], [0045]-[0046]), wherein a parameterized function ascertains the probability of drawing each subsequent edge based on the subpath drawn so far through the directed graph, and wherein the parameterization of the parameterized function is adjusted during the training with respect to the cost function (controller receives an embedding of the preceding action and can use recurrent layers to retain information about earlier decisions)([0056]-[0059],[0063])..(update controller parameters using reinforcement learning and expressly identifies policy-gradient optimization) ([0078]–[0080]); and
drawing a path depending on the adjusted probabilities and creating the machine learning system corresponding to the drawn path (generating a new path using the trained controller parameters and using the architecture defined by that path as the final architecture…outputting the resulting architecture and, in some embodiments, the trained network parameters)([0047]-[0049]).
Houlsby does not explicitly disclose each subsequent edge is assigned a probability based on which the respective edge is drawn, wherein the probabilities are ascertained depending on a sequence of previously drawn edges of the respective path, such that the probability of drawing each subsequent edge is based on a subpath drawn so far through the directed graph.
However, Benyahia discloses a recurrent policy that produces architectural actions sequentially and conditions the current action on preceding actions ([0112]-[0113]). Therefore, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Houlsby with Benyahia features. One would have been motivated to do so to allow later architecture decisions to account for the previously traversed path.
Claim 13. Houlsby and Benyahia disclose the method according to claim 11, Houlsby further discloses wherein the previously drawn edges and/or nodes of the subpath are assigned a unique coding of their order (numerical embeddings of the preceding action and the current decision point, generated using embedding tables.. processes action and state embeddings through its controller to generate the outgoing-edge score distribution) ([0056]-[0064]) and Benyahia further discloses the parameterized function ascertains the probabilities depending on the coding (an architecture as an ordered sequence of actions…history-conditioned stochastic policy) ([0112]-[0113]). One would have been motivated to do so to allow later architecture decisions to account for the previously traversed path.
Claim 14. Houlsby and Benyahia disclose the method according to claim 11, Houlsby further discloses wherein the parameterized function ascertains a probability distribution over possible edges, from a set of edges that can be drawn next (generating a score distribution containing a score for each outgoing edge and sampling an action from that distribution. Where the scores are probabilities, sampling occurs directly from the probability distribution)([0056], [0084]-[0085]).
Claim 15. Houlsby and Benyahia disclose the method of claim 11, Houlsby further discloses wherein the parameterized function is an affine transformation or a neural network (..controller as a neural network with trainable parameters and describes feedforward and recurrent layers)([0040],[0060]-[0064]).
Claim 16. Houlsby and Benyahia disclose the method according to claim 13, Houlsby further discloses wherein the parameterized function is an affine transformation or a neural network, and wherein the parameterization of the affine transformation describes a linear transformation and a shift of the unique coding, and a scaling is composed of a low-rank approximation and the scaling depending on a number of edges (..controller as a neural network with trainable parameters and describes feedforward and recurrent layers)([0040],[0060]-[0064]). [The neural-network alternative is expressly disclosed; a separate affine-transformation teaching is unnecessary to address a claim written in the alternative.]
Claim 17. Houlsby and Benyahia disclose the method according to claim 15, Houlsby further discloses wherein a plurality of parameterized functions are used and the parameterized functions are each provided by a neural network (multiple node-specific neural networks that generate score distributions for the outgoing edges of their respective graph nodes)(0060]-[0062], [0064]), wherein a parameterization of a plurality of layers of the neural networks are shared among all neural networks (a common neural-network component shared across graph nodes, including an embodiment having feedforward and recurrent layers)([0060]-[0063]).
Claims 18-19 represent the storage and apparatus of claim 11, respectively and are rejected along the same rationale.
Response to Arguments
6. Applicant’s arguments and amendments filed on 07/21/2026 have been fully considered
but are moot in view of new ground of rejection(s).
Conclusion
7. Applicants’ amendment necessitated the new ground(s) 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.
8. The prior art made of record and not relied upon is considered pertinent to applicant's disclosure (See PTO-892).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Phenuel S. Salomon whose telephone number is (571) 270-1699. The examiner can normally be reached on Mon-Fri 7:00 A.M. to 4:00 P.M. (Alternate Friday Off) EST.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Usmaan Saeed can be reached on (571) 272-4046. The fax phone number for the organization where this application or proceeding is assigned is 571-273-3800.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/PHENUEL S SALOMON/Primary Examiner, Art Unit 2146
