[Paper Review] Dependency Injection for Programming by Optimization

Zoltan A. Kocsis, Jerry Swan|arXiv (Cornell University)|Jul 13, 2017

Evolutionary Algorithms and Applications18 references3 citations

TL;DR

This paper introduces ContainAnt, a Dependency Injection-based framework that replaces traditional weaver tools in Programming by Optimization (PbO), enabling efficient, on-line algorithm configuration. By modeling component wiring as grammatical optimization over context-free grammars, it introduces novel evolutionary heuristics—GrAnt and GrEvo—that outperform SMAC by up to two orders of magnitude in optimization speed and solution quality.

ABSTRACT

Programming by Optimization tools perform automatic software configuration according to the specification supplied by a software developer. Developers specify design spaces for program components, and the onerous task of determining which configuration best suits a given use case is determined using automated analysis tools and optimization heuristics. However, in current approaches to Programming by Optimization, design space specification and exploration relies on external configuration algorithms, executable wrappers and fragile, preprocessed programming language extensions. Here we show that the architectural pattern of Dependency Injection provides a superior alternative to the traditional Programming by Optimization pipeline. We demonstrate that configuration tools based on Dependency Injection fit naturally into the software development process, while requiring less overhead than current wrapper-based mechanisms. Furthermore, the structural correspondence between Dependency Injection and context-free grammars yields a new class of evolutionary metaheuristics for automated algorithm configuration. We found that the new heuristics significantly outperform existing configuration algorithms on many problems of interest (in one case by two orders of magnitude). We anticipate that these developments will make Programming by Optimization immediately applicable to a large number of enterprise software projects.

Motivation & Objective

• To address the limitations of existing weaver-based tools in Programming by Optimization (PbO), which rely on fragile language extensions and external command-line configuration.
• To enable seamless integration of PbO into existing software projects by replacing syntactic extensions with Dependency Injection containers.
• To exploit the structural correspondence between Dependency Injection and context-free grammars to derive a new class of evolutionary metaheuristics for automated algorithm configuration.
• To evaluate the performance of these new heuristics against established tools like SMAC on real-world optimization problems.
• To demonstrate that the proposed approach enables faster, more reliable, and more scalable optimization than current PbO pipelines.

Proposed method

• Represents software component design spaces as context-free grammars, where each production rule corresponds to a configuration choice in the dependency graph.
• Uses Dependency Injection containers to instantiate component configurations without modifying source code, enabling runtime and on-line optimization.
• Applies a novel grammar-based genetic algorithm (GrAnt) and ant colony optimization (GrEvo) to explore the design space by evolving valid grammatical derivations.
• Maps the optimization problem to a search over derivations in a context-free grammar, where fitness is determined by performance on a given benchmark.
• Employs a configuration-aware mapping from abstract syntax trees to concrete component instances, preserving type safety and modularity.
• Introduces a heuristic-based search strategy that avoids the need for external weavers or command-line wrappers, enabling real-time configuration.

Experimental results

Research questions

• RQ1 Can Dependency Injection be used as a native, lightweight alternative to syntactic weavers in Programming by Optimization?
• RQ2 Does the structural isomorphism between Dependency Injection and context-free grammars enable effective metaheuristic search for algorithm configuration?
• RQ3 How do grammar-based evolutionary heuristics (GrAnt and GrEvo) compare in performance and robustness to established tools like SMAC on real-world optimization problems?
• RQ4 Can the proposed approach achieve significant speedups and solution quality improvements over SMAC in practical enterprise software scenarios?
• RQ5 What are the limitations of existing grammatical evolution approaches (e.g., early convergence) when applied to software configuration?

Key findings

• The GrAnt heuristic significantly outperformed all other methods on four out of five benchmark problems, achieving globally optimal solutions on the Branin function and both subset sum instances.
• On the Branin function, GrAnt achieved an average fitness of 0.87, compared to SMAC’s 1.47, with a p-value < 0.001, indicating statistically significant improvement.
• For the subset sum P02 problem, GrAnt achieved an average fitness of 0.65, outperforming SMAC’s 1.47 and random search’s 0.03, with a p-value < 0.001.
• In the DHeap problem, GrAnt achieved a mean fitness of 46,801 with zero variance, outperforming SMAC’s 46,752 and random search’s 46,594.
• The optimization time for a four-hour SMAC task was reduced to 46 seconds using GrAnt, representing a 50x speedup while improving solution quality.
• GrEvo showed poor performance due to early loss of diversity, converging in as few as five generations, suggesting limitations in standard grammatical evolution for real-time software configuration.