George Forman (no, not that George Foreman) invented a method for selecting a predetermined number of features for a set of binary partitions over a set of categories given a dataset of feature vectors associated with the categories. Read the application here. Claim 1 and 4:
1. A computer-implemented feature selection method for selecting a predetermined number of features for a set of binary partitions over a set of categories given a dataset of feature vectors associated with the categories, the method comprising:for each binary partition under consideration, ranking features using two-category feature ranking; andwhile the predetermined number of features has not yet been selected:picking a binary partition p;selecting a feature based on the ranking for binary partition p; andadding the selected feature to an output list if not already present in the output list and removing the selected feature from further consideration for the binary partition p.4. A method in accordance with claim 1 and further comprising using the selected features in training a classifier for classifying data into categories.
As the U.S. Supreme Court has noted, “‘[a]n idea of itself is not patentable’….‘A principle, in the abstract, is a fundamental truth; an original cause; a motive; these cannot be patented, as no one can claim in either of them an exclusive right.’” [Diamond v. Diehr, 450 U.S. 175, 185 (1981)] (citations omitted).In determining whether a claim as a whole is directed to an abstract idea, a key distinction has been drawn between (1) claims that seek to wholly pre-empt the use of a fundamental principle, and (2) claims that are merely limited to foreclosing others from using a particular application of that fundamental principle. See In re Bilski, 545 F.3d 943, 957 (Fed. Cir. 2008) (en banc), cert. granted, 77 U.S.L.W. 3442, 3653, 3656 (U.S. June 1, 2009) (No. 08-964).Based on U.S. Supreme Court precedents, Bilski restated the U.S. Supreme Court’s “definitive test to determine whether a process claim is tailored narrowly enough to encompass only a particular application of a fundamental principle rather than to pre-empt the principle itself.” Bilski, 545 F.3d at 954. This restatement, embodied as the “machine-or-transformation test,” requires that a claimed process either (1) be tied to a particular machine or apparatus, or (2) transform a particular article into a different state or thing. Id. This test ensures that the claimed process does not pre-empt uses of the principle that do not use the specified machine or apparatus. The test further precludes a claimed process from pre-empting “the use of the principle to transform any other article, to transform the same article but in a manner not covered by the claim, or to do anything other than transform the specified article.” Id.
A classifier is typically constructed using an inducer. An inducer is an algorithm that builds the classifier using a training set comprising records with labels. After the classifier is built, it can be used to classify unlabeled records.” Spec. ¶ 0005.
A “classifier” is therefore a particular machine in that it performs a particular data classification function that is beyond mere general purpose computing. Since the claim recites using the selected features in training this classifier, the claim is therefore tied to a particular machine, namely the classifier. Claim 4 also transforms a particular article into a different state or thing, namely by transforming an untrained classifier into a trained classifier.
If a claimed machine (or article of manufacture) involves a mathematical algorithm, then we must determine whether the scope of the claimed invention encompasses one of the judicially-created exceptions. This determination of claim scope requires that we make two inquiries:(1) Is the claim limited to a tangible practical application, in which the mathematical algorithm is applied, that results in a real-world use (e.g., “not a mere field-of-use label having no significance”)?(2) Is the claim limited so as to not encompass substantially all practical applications of the mathematical algorithm either “in all fields” of use of the algorithm or even in “only one field?”If the machine (or article of manufacture) claim involves a mathematical algorithm and fails either prong of this two-part inquiry, then the claim is not directed to patent-eligible subject matter under § 101.
Here, claim 15 is not limited to a tangible practical application in which the algorithm is applied that results in a real-world use. Although the preamble recites that the program code is “for feature selection for a plurality of categories,” this nominal intended use recitation falls well short of reciting a tangible practical application that results in a real-world use. Moreover, the remainder of the claim merely recites data manipulation steps that are performed by the processor upon execution of the program code. These limitations, likewise, fail to recite a tangible practical application that results in a real-world use. Furthermore, the sheer breadth of claim 15 encompasses substantially all practical applications of the algorithm. In essence, claim 15 merely recites a general purpose computing device intended to facilitate the future execution of the recited data manipulation functions—manipulations similar to those in claim 1 that we found to be unpatentable under § 101.