The Ongoing Patent Dispute Over Innovative ML-Based Pattern Recognition

• Published date: 20 April 2022
• Subject Matter: Intellectual Property, Technology, Patent, New Technology
• Law Firm: Oblon, McClelland, Maier & Neustadt, L.L.P
• Author: Ms Yuki Onoe

Support Vector Machine-Recursive Feature Elimination (SVM-RFE) is a technology which can be used to find relevant patterns in a large data set such as the data generated in the sequencing of genomes and produce smaller subsets. In Health Discovery Corp. v. Intel Corp.¹, the patent owner HDC, in its complaint for infringement, discussed the innovative aspects of the technology:

Support Vector Machine ' Recursive Feature Elimination ('SVM-RFE') is an application of SVM that was invented by Dr. Weston and Dr. Guyon as members of HDC's science team, to find discriminate relationships within clinical datasets, as well as within gene expression and proteomic datasets created from micro-arrays of tumor versus normal tissues. In general, SVMs identify patterns ' for instance, a biomarker/genetic expression signature of a disease. The SVM-RFE utilizes this pattern recognition capability to identify, rank and order the features that contribute most to the desired results, and successively eliminate the features with the lowest rank order, until the optimal feature set is obtained to define the model.

However, Judge Albright, in his December 27, 2021 opinion, stated that the patent claim reciting the pattern recognition method would 'merely improve or 'enhance' an abstract idea'² and satisfy Alice step one, meaning it is directed to judicial exception of abstract idea.

Judge Albright analyzed whether the claim is directed to a 'specific means or method that improves [that] relevant technology.'³ The claim would be found eligible in the Alice step one if it is directed to 'improvements to the functioning of a computer or network.'⁴ However, looking at representative claim 1 (below) of U.S. Patent No. 7,177,188 (the 188 patent), Judge Albright stated that 'the claims here merely produce data with improved quality relative to that produced by conventional mathematical methods.'⁵ The 'relevant technology' that is improved is an abstract, mathematical method, and the improvement is not tied to the 'physical,'⁶ which was the distinction over the cases such as McRO where the improvement was 'allowing computers to produce 'accurate and realistic lip synchronization and facial expressions in animated characters.''⁷

1. A computer-implemented method for identifying patterns in data,

the method comprising:

(a) inputting into at least one support vector machine of a plurality

of support vector machines a training set having known outcomes,

the at least one support vector...