Ex parte Desjardins

When a claim recites an abstract idea (here, a mathematical calculation) at Step 2A, Prong One, but the claim as a whole reflects an improvement to how a machine learning model itself operates, as disclosed in the specification, the claim integrates the exception into a practical application at Prong Two. It is therefore not directed to an abstract idea, and the Board cannot sustain a § 101 rejection by treating machine learning as an unpatentable algorithm and the remaining elements as generic computer components. Examiners and panels must not evaluate such claims at a high level of generality. No prior Board precedential or informative decision is named in the text. The Appeals Review Panel vacated the Board's own new ground of rejection, which it faulted as overbroad and as a cursory analysis that ignored Enfish. For examiners and panels, the decision means a § 101 rejection of machine learning claims needs a real Prong Two analysis of technological improvement, and the ARP says §§ 102, 103 and 112 should be the focus of examination. For applicants and appellants, a specification that identifies improvements to model training, such as learning new tasks while protecting earlier ones and using less storage, can support eligibility if the claim reflects the improvement. The ARP relied on Alice, Enfish (software can make non-abstract improvements to computer technology), and Intellectual Ventures I LLC v. Symantec Corp. (a specification's assertion alone is not enough without a claim reflecting the improvement). LEGAL TEST OR STANDARD: The ARP applied the MPEP Step 2A framework consistent with Alice step one. Prong One asks "does the claim recite an abstract idea, law of nature, or natural phenomenon?" Prong Two asks "does the claim recite additional elements that integrate the judicial exception into a practical application?" The Prong Two inquiry evaluates the significance of the additional elements relative to the invention, with the ultimate question being whether the exception is integrated into a practical application. If the claim as a whole is not directed to a judicial exception at Prong Two, the eligibility analysis ends. Improvements to the functioning of a computer or to other technology or technical field count (MPEP 2106.04(d)(1), 2106.05(a)), while generally linking a judicial exception to a technological environment does not (MPEP 2106.05(h)). TRIGGERING FACTS: Claim 1 recited computing an approximation of a posterior distribution, which the Board found to be a mathematical calculation, and the appellant did not dispute that finding. The appellant argued technical improvements in continual learning and model efficiency, citing Spec. paragraph 21: learning new tasks in succession while protecting knowledge of earlier tasks, less storage, and reduced system complexity. The ARP found the claim itself reflected that improvement through the limitation to adjust the first values of the parameters to optimize performance on the second task while protecting performance on the first task. The Board had entered the section 101 rejection sua sponte as a new ground, with only a conclusory view that the remaining elements were generic computer components. KEY QUOTE: "We are persuaded that constitutes an improvement to how the machine learning model itself operates, and not, for example, the identified mathematical calculation." (p. 9) PRIOR PRECEDENT AFFECTED: No prior precedential or informative decision is directly affected. The ARP vacated the Board's own Decision on Appeal new ground of rejection under 35 U.S.C. 101 in this appeal, and followed Enfish, LLC v. Microsoft Corp., 822 F.3d 1327 (Fed. Cir. 2016). PRACTICE TIP: When defending machine learning or other software claims against a section 101 rejection, concede nothing on Prong Two. Identify the specific technical improvement in the specification, such as better model training, less storage or reduced complexity, and point to the claim limitation that reflects it, citing Enfish and MPEP 2106.04(d)(1) and 2106.05(a).