Multi-Objective Optimization Using Composite Novelty Selection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Evolutionary algorithms face challenges in multi-objective optimization problems, where they often struggle to find diverse solutions that optimize multiple objectives simultaneously without losing diversity in the search space, especially when objectives are not zero-sum alternatives.

Innovation Solution

A multi-objective driven algorithm that uses composite functions to focus the search in more useful areas of the search space, incorporating a dominance filter and relative novelty and diversity measures to select optimal individuals, ensuring a diverse group of individuals is produced without losing diversity.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If evolutionary algorithms use multiple objectives to optimize different aspects of solutions, then the ability to find trade-off solutions improves, but the difficulty of maintaining diversity in the search space increases

Engineering Contradiction:

Improveability to optimize multiple objectivesVSAvoiddifficulty of maintaining diversity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent segments the selection process into multiple stages: first applying a dominance filter to identify non-dominated individuals, then applying novelty and diversity measures to select from that filtered set. This segmentation allows the algorithm to handle multiple objectives systematically while maintaining diversity through staged filtering rather than attempting to optimize all objectives simultaneously in a single step.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent changes the selection parameters by introducing composite functions that combine multiple objectives with adjustable weights, and by using novelty and diversity measures as additional selection criteria. These parameter changes allow the algorithm to adapt the optimization process to maintain diversity while optimizing multiple objectives, transforming the selection mechanism from simple fitness-based to a multi-criteria decision process.

Inventive Principle:

Principle #35Parameter changes

2Productivity

If the algorithm focuses search in specific areas using composite functions, then search efficiency improves, but the risk of losing diversity in the search space increases

Engineering Contradiction:

Improvesearch efficiencyVSAvoiddiversity of solutions

Core Design Contradiction:

ProductivityVSQuantity of substance

Solution Approach 1:

The patent incorporates feedback mechanisms by calculating novelty and diversity measures for individuals in the focused search areas, then using these measures to adjust selection decisions. This feedback loop ensures that even when the search is focused on specific areas through composite functions, the algorithm continuously monitors and maintains diversity by selecting individuals that contribute to the overall diversity of the population.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent adds another dimension to the selection process by incorporating novelty and diversity measures as additional criteria beyond the composite function values. This dimensional expansion allows the algorithm to focus search efficiently using composite functions while simultaneously maintaining diversity through the additional selection dimension, effectively adding a diversity-preserving layer to the focused search mechanism.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

3Speed

If the algorithm selects only the most fit individuals, then convergence speed improves, but the ability to explore new areas of the search space deteriorates

Engineering Contradiction:

Improveconvergence speedVSAvoidability to explore new areas

Core Design Contradiction:

SpeedVSAdaptability or versatility

Solution Approach 1:

The patent applies preliminary filtering through the dominance filter before applying novelty and diversity measures. This preliminary action identifies non-dominated individuals that represent promising areas of the search space, allowing the algorithm to converge quickly on high-quality solutions while the subsequent novelty and diversity measures ensure that individuals representing new or under-explored areas are not prematurely eliminated.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent applies partial selection by not selecting all non-dominated individuals, but only those that also meet novelty and diversity criteria. This partial action prevents premature convergence by deliberately selecting some individuals based on diversity considerations rather than purely on fitness, allowing the algorithm to maintain exploration capability while achieving convergence on promising solutions.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS11527308B2Enhanced optimization with composite objectives and novelty-diversity selection

Publication Date: 2022.12.13 COGNIZANT TECHNOLOGY SOLUTIONS US CORP

AI summary

A composite novelty method approach to deceptive problems where a secondary objective is available to diversify the search is described. In such cases, composite objectives focus the search on the most useful tradeoffs and allow escaping deceptive areas. Novelty-based selection increases exploration in the focus area, leading to better solutions, faster and more consistently and it can be combined with other fitness-based methods.