The application of the Alice/Mayo two-step framework to mathematical neural network optimization algorithms and tensor transformation claims.
Under 35 U.S.C. § 101 and the two-step Alice/Mayo framework, judicial exceptions bar patenting laws of nature, natural phenomena, and abstract mathematical ideas. Step 1 asks whether the claim is directed to an abstract concept (such as matrix multiplication or gradient descent). If so, Step 2 examines whether the claim recites an 'inventive concept'—significantly more than routine, conventional computer implementation. Patenting pure neural net structures without physical or system improvements routinely fails § 101.
Federal Circuit jurisprudence rigorously strikes down claims that recite conventional algorithms executed on generic hardware, while upholding patents that improve internal computer efficiency, data compression, or sensor physical operation.
