Population-Based Exhaustive Replacement for Variable Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional variable selection methods for generating predictive models, such as those used in the pharmaceutical industry for drug discovery, are computationally intensive and often fail to provide optimal solutions due to slower convergence and lower predictive capabilities, especially when dealing with larger datasets.
Innovation Solution
A processor-implemented population-based exhaustive replacement method that iteratively selects and replaces variables to optimize predictive models, using predefined criteria to ensure optimal subset selection and parallel processing to reduce computational time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If exhaustive search approach is used to evaluate each possible combination of variables, then optimal solution is provided, but computational complexity increases significantly limiting applications to smaller subsets
Solution Approach 1:
The exhaustive search space is segmented into multiple populations of variable subsets. Instead of evaluating all possible combinations at once, the method divides them into manageable groups (populations) that are processed iteratively. Each population undergoes replacement operations where variables are systematically substituted and evaluated, allowing the computational burden to be distributed across multiple smaller evaluation cycles rather than one overwhelming exhaustive search.
Solution Approach 2:
The method performs preliminary actions by initializing multiple populations with diverse variable subsets before the main optimization process. These initial populations are prepared with specific structures and properties that guide subsequent replacement operations. By pre-organizing the search space into structured populations rather than random combinations, the method reduces the effective complexity of the exhaustive search while maintaining solution optimality.
2Productivity
If random search techniques such as genetic algorithm or particle swarm optimization are used, then computational intensity is reduced, but guarantee of optimal solution is lost
Solution Approach 1:
The method introduces dynamics by implementing iterative replacement operations within populations. Rather than using static random search, the algorithm dynamically evolves variable subsets through systematic replacement of variables with alternatives from the same population. This dynamic evolution process allows the method to maintain computational efficiency like random search while converging toward optimal solutions through structured exploration of the variable space.
Solution Approach 2:
The method incorporates feedback mechanisms by evaluating objective functions for each modified subset and using these evaluations to guide subsequent replacement decisions. The performance metrics from each population iteration feed back into the selection process for the next iteration, allowing the algorithm to learn from previous evaluations and progressively improve solution quality. This feedback loop ensures convergence toward optimal solutions while maintaining computational efficiency.
3Measurement precision
If conventional replacement method is used to select descriptors iteratively, then variable selection is achieved, but convergence speed is slow and predictive capability is limited
Solution Approach 1:
The method merges multiple conventional replacement method iterations into parallel population-based processing. Instead of sequentially replacing variables one at a time across single descriptor sets, the approach combines multiple descriptor sets into populations and performs replacements across all populations simultaneously. This merging of operations into parallel structures accelerates convergence by processing multiple variable selection paths concurrently while maintaining the accuracy benefits of iterative replacement.
Solution Approach 2:
The method adds a population dimension to the traditional variable selection process. Rather than working with a single sequence of variable subsets, the approach introduces multiple populations as an additional dimension of exploration. This dimensional expansion allows simultaneous exploration of multiple variable selection trajectories, reducing the time to converge on optimal subsets by distributing the search effort across parallel dimensions rather than sequential steps.
Data Source
AI summary
Population based exhaustive replacement method(s) (PERM) for optimal variables selection and generation of regression models, to overcome conventional approaches, thereof is described herein. PERM initializes population based on one or more criteria, wherein one or more paths for variables/descriptors 1 to r for replacement with remaining descriptors wherein the one or more paths are updated based on relative error associated with each variable. For each combination of descriptors of r size, inter correlation of the descriptors are verified and predictive models are built. Subsets of variable with higher predictive ability are selected for substitution of initial population to obtain an updated population on which a replacement method is performed to obtain optimal set of variables. One or more variables of optimal set are randomly replaced, and perturbation can be performed on the top best subsets to converge at global optimum.