Partial Inversion of Machine Learning Models for Proactive Input Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine learning models are unable to proactively indicate the ranges of input values that will result in a specific output value, limiting their ability to prevent or achieve a desired outcome in systems like web service outages.
Innovation Solution
The implementation of partial inversion of machine learning models, which allows specifying a target output value and determining the corresponding input values to achieve or avoid that output, enabling proactive decision-making in systems like remote network management platforms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional machine learning models are used to predict output values from input values, then prediction accuracy is improved, but the ability to proactively determine input ranges for desired outputs deteriorates
Solution Approach 1:
The patent applies inversion by transforming the conventional machine learning prediction direction. Instead of predicting output from input (forward direction), the system inverts the model to determine acceptable input ranges that will produce a desired output (backward direction). This is achieved by inverting the trained model f(x) to solve for x given y, enabling proactive determination of input parameters that will achieve target outcomes while maintaining the original model's prediction accuracy
2Reliability
If machine learning models are trained with large data sets to improve prediction accuracy, then model reliability is improved, but computational complexity increases
Solution Approach 1:
The patent applies preliminary action by performing the computationally intensive model training with large data sets once during the offline phase to establish a reliable trained model. The inverted model then reuses this pre-trained model for multiple proactive queries without requiring retraining. This approach maintains high model reliability through comprehensive training while reducing computational complexity during operational use, as the inversion process leverages the existing trained model rather than requiring additional training computations
Data Source
AI summary
An example embodiment may involve a machine learning model representing relationships between a dependent variable and a plurality of n independent variables. The dependent variable may be a function of the n independent variables, where the n independent variables are measurable characteristics of computing devices, and where the dependent variable is a predicted behavior of the computing devices. The embodiment may also involve obtaining a target value of the dependent variable, and separating the n independent variables into n−1 independent variables with fixed values and a particular independent variable with an unfixed value. The embodiment may also involve performing a partial inversion of the function to produce a value of the particular independent variable such that, when the function is applied to the value of the particular independent variable and the n−1 independent variables with fixed values, the dependent variable is within a pre-defined range of the target value.