Ant-based metaheuristics have successfully been applied to a variety of different graph-based problems. However, for Cartesian Genetic Programming (CGP) only the impact of Max-Min Ant Systems has been tested. In this work, we try to fill this gap by applying four different popular ant-based metaheuristics as the optimizer (and therefore training algorithm) of CGP. The idea of combining CGP with ant-based metaheuristics is not novel but older works’ experimental design may not meet today’s standard. To compare these metaheuristics to the Evolution Strategies (ESs) commonly used in CGP, we benchmark against a standard CGP variant that uses a simplistic (1+4) -ES, mutation, and no crossover. Additionally, we include (μ +λ ) -ES and (μ , λ ) -ES in our experiments. We analyse the performance on datasets from the symbolic regression, regression, and classification domains. By tuning and evaluating various configurations, we can not affirm a significant improvement by using ant-based methods with CGP as we encounter premature convergence—even with those ant-based metaheuristics that were originally proposed to overcome such problems. Despite our results being of negative nature, this work still gives important and interesting insights into the training of CGP models. The key contributions of our work are thus a more thorough benchmarking of these optimizers than has been done before. This should clear up doubts about the capabilities of ant-based metaheuristics in CGP. Furthermore, we include a roadmap on how they can be addressed to solve this complex optimization problem from the model building domain of machine learning.
The container relocation problem is an important combinatorial optimisation problem commonly found in warehouses and container ports. The goal of this problem is to retrieve all of the containers from the yard with the fewest container relocations between the stacks. Since the problem is NP-hard, various heuristics have been proposed to solve it, among which are relocation rules (RRs), simple constructive heuristics that incrementally construct the solution. However, it is quite difficult to design such RRs manually, so genetic programming has been applied to design new RRs automatically. A significant problem with RRs, whether manually or automatically designed, is that they usually have a limited view of the problem. This means that they will often make decisions that can negatively influence the future, meaning that the current decision would cause additional relocations. Therefore, this study investigates different relocation schemes that can be used within RRs to obtain rules with lookahead ability. These rules will enable containers to be relocated based on future information and, consequently, arranged better in the yard. For that purpose, three novel relocation schemes for automatically designed RRs are defined and evaluated on an existing problem set. The results demonstrate that integrating additional elements to evolve lookahead RRs can significantly improve the results.
The Semantic Learning algorithm based on Inflate and deflate Mutation (SLIM) is a promising recent variant of Geometric Semantic Genetic Programming (GSGP) that introduces a new Deflate Geometric Semantic Mutation (DGSM). This operator maintains the key feature of the standard Geometric Semantic Mutation (GSM), inducing a unimodal error surface for any supervised learning problem, while generating smaller offspring than their parents, and thus allowing SLIM to generate compact, and potentially interpretable, final solutions. A key parameter controlling the evolution process in both GSGP and SLIM is the Mutation Step (MS), which regulates the extent of perturbation to the parent semantics. While it is intuitive that the optimal value of MS has a relationship with the scale of the dataset features, to the best of our knowledge no prior research has extensively explored this relationship. In this work, we provide the first comprehensive investigation into this topic. First, we hypothesize a general rule by analyzing results from artificial datasets, and then we confirm these findings with more complex, real-world datasets. This approach offers a solid alternative to the typical hyperparameter tuning approach.
This article presents an interpretable approach to binary image classification using Genetic Programming (GP), applied to the Patch-Camelyon (PCAM) dataset, which contains small tissue biopsy patches labeled as malignant or benign. While Deep Neural Networks (DNNs) achieve high performance in image classification, their opaque decision-making processes, prone to overfitting behavior and dependency on large amounts of annotated data limit their utility in critical fields like digital pathology, where interpretability is essential. To address this, we employ GP, specifically using the Multi-Modal Adaptive Graph Evolution (MAGE) framework, to evolve end-to-end image classification pipelines. We trained MAGE a hundred times with the best optimized key hyperparameters for this task. Among all MAGE models trained, the best one achieved 78
This study investigates the effectiveness of Cartesian Genetic Programming by analyzing numerous indicators of evolutionary dynamics when using different crossover operators and the canonical mutation-only ( 1+4 ) strategy. Specifically, we examine a traditional crossover operator which is based on the random selection of parental genes; Subgraph Crossover, where points in the range of active nodes are considered; and the recently-proposed Deep Neural Crossover (DNC) approach which utilizes a transformer network to learn correlations between genes and predict potentially beneficial crossover points. The performance of these different crossovers is evaluated on 11 standard and one real-world regression problem.
