PTAB

IPR2026-00223

Strategy Inc v. Kelly D Phillipps

Log In
Key Events
Petition

1. Case Identification

2. Patent Overview

  • Title: USER INTERFACE FOR MACHINE LEARNING
  • Brief Description: The ’574 patent discloses systems and methods for a user interface that presents machine learning results. The technology centers on pre-computing and storing permutations of machine learning results, allowing a user to adjust input parameters and view the corresponding predicted outcomes with little or no processing delay.

3. Grounds for Unpatentability

Ground 1A: Claims 1-3, 5, 8-10, 12, and 13 are obvious over Johnson and Lin.

  • Prior Art Relied Upon: Johnson (Application # 2004/0138932) and Lin (Patent 8,364,613).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner argued that Johnson taught a "digital cockpit" user interface for business forecasting that pre-calculates and caches results to provide instantaneous "what-if" analysis. This system disclosed the core concepts of an input module for receiving user-defined parameters, a pre-compute module ("pre-loading logic"), a display module, and an update module that dynamically shows results from the cache as a user adjusts inputs. Petitioner contended that Lin taught the missing element: a "predictive compiler module." Lin disclosed training machine learning models from data, combining them into more accurate ensembles, and implementing these trained models as executable computer programs, which Petitioner asserted is the function of a compiler in this context.
    • Motivation to Combine: A POSITA would combine Lin's methods with Johnson's system to improve its predictive capabilities. Johnson described using predictive models but did not detail how they were trained. A POSITA would have looked to a known technique like Lin's to train more accurate models, combine them into ensembles for better predictions, and implement them as efficient, executable programs, thereby enhancing Johnson's forecasting system.
    • Expectation of Success: A POSITA would have a high expectation of success, as this combination involved applying Lin’s known machine learning training techniques to improve Johnson’s forecasting system, which was already designed to use such predictive models.

Ground 1B: Claims 7, 11, and 16-18 are obvious over Johnson, Lin, and Purcell.

  • Prior Art Relied Upon: Johnson (Application # 2004/0138932), Lin (Patent 8,364,613), and Purcell (Patent 5,727,161).
  • Core Argument for this Ground:
    • Prior Art Mapping: This ground builds upon the Johnson/Lin combination by adding the teachings of Purcell. Petitioner argued Purcell taught a user interface where a user can adjust a desired output or "goal value" (e.g., via a vertical scroll bar), and the system, in response, displays the required input values needed to achieve that goal. This directly addressed limitations in claims like claim 7, which requires machine learning parameters to comprise both an input adjusted by the user and an output adjusted by the user. Purcell also taught displaying results in a spreadsheet or graphical table format, as required by claim 11.
    • Motivation to Combine: A POSITA would be motivated to incorporate Purcell's goal-seeking interface into the Johnson/Lin system to enhance its "what-if" analysis functionality. This would allow users not just to see the results of changing inputs, but also to more easily determine the specific inputs required to achieve a desired business outcome, a clear improvement for a business forecasting tool.

Ground 1C: Claims 21-25 are obvious over Johnson, Lin, and Mihaylov.

  • Prior Art Relied Upon: Johnson (Application # 2004/0138932), Lin (Patent 8,364,613), and Mihaylov (Application # 2013/0024160).

  • Core Argument for this Ground:

    • Prior Art Mapping: This ground builds on the Johnson/Lin combination by adding the teachings of Mihaylov to address limitations related to a "collaboration module" in independent claim 21. Petitioner asserted that Mihaylov taught techniques to analyze and display the sources of error in predictive forecasts, including calculating the "model change impact"—the difference in accuracy resulting from using a different predictive model. Petitioner mapped this functionality to the claimed "collaboration module," which determines an "impact on machine learning results from a different one of the predictive programs."
    • Motivation to Combine: A POSITA would combine Mihaylov's error analysis with the Johnson/Lin system to meet Johnson's stated goal of allowing users to "assess the accuracy" of forecasts. By incorporating Mihaylov's teachings, the system could provide users with a sophisticated understanding of model performance and the specific impact of model changes, which is a logical and predictable improvement for a financial forecasting platform.

  • Additional Grounds: Petitioner asserted an additional obviousness challenge (Ground 1D) for claims 4, 6, 14, 15, 19, and 20 based on the combination of all four references: Johnson, Lin, Purcell, and Mihaylov.

4. Key Claim Construction Positions

  • Petitioner argued that several claim terms reciting a "module" (e.g., "predictive compiler module," "pre-compute module," "collaboration module") are means-plus-function terms under 35 U.S.C. §112(f), despite lacking the phrase "means for."
  • Citing Williamson v. Citrix Online, LLC, Petitioner contended that "module" is a nonce word that can operate as a substitute for "means" and that the claims recite these modules in purely functional terms without sufficient corresponding structure.
  • Petitioner identified the algorithms and structures described in the ’574 patent's specification that correspond to the functions of each "module" to construe the scope of these means-plus-function limitations for its invalidity analysis.

5. Relief Requested

  • Petitioner requests institution of an inter partes review and cancellation of claims 1-25 of the ’574 patent as unpatentable.