Genetic Programming. 26th European Conference, EuroGP 2023 Held as Part of EvoStar 2023 Brno, Czech Republic, April 12–14, 2023 Proceedings

Description

Preface Organization Contents Long Presentations A Self-Adaptive Approach to Exploit Topological Properties of Different GAs' Crossover Operators 1 Introduction 2 Fundamental Concepts 2.1 Crossover 2.2 Convex Combination, Convex Hull, and Convex Search 3 Related Works 4 Methodology 4.1 Dynamic Diversity Maintenance 4.2 Self-adaptive Crossover 5 Experimental Settings 6 Experimental Results 7 Conclusions References A Genetic Programming Encoder for Increasing Autoencoder Interpretability 1 Introduction 1.1 Structure 2 Background and Related Work 2.1 Non-linear Dimensionality Reduction 2.2 Evolutionary Computation for Dimensionality Reduction 2.3 Genetic Programming for Autoencoding 3 Proposed Method: GPE-AE 3.1 GP Representation of Encoder 3.2 Fitness Evaluation 3.3 Decoder Architecture 4 Experiment Design 4.1 Comparison Methods 4.2 Evaluation Measures 4.3 Datasets 5 Results 6 Further Analysis 7 Conclusions References Graph Networks as Inductive Bias for Genetic Programming: Symbolic Models for Particle-Laden Flows 1 Introduction 2 Background and Related Work 2.1 Genetic Programming in Physics Applications 2.2 Machine Learning for Particle-Laden Flows 3 Proposed Methods 3.1 Graph Networks 3.2 Genetic Programming 4 Experiment Design 4.1 Data Generation: Simulation of Particle-Laden Flows 4.2 Data Preprocessing 4.3 Algorithm Settings 5 Results and Analysis 5.1 Overall Algorithm Performance 5.2 Explainability of Equations 5.3 Validation of Symbolic Models 6 Conclusion and Future Work References Phenotype Search Trajectory Networks for Linear Genetic Programming 1 Introduction 2 The LGP System 2.1 Boolean LGP Algorithm 2.2 Genotype, Phenotype, and Fitness 3 Kolmogorov Complexity 4 Sampling and Metrics Estimation 5 Search Trajectory Networks 5.1 General Definitions 5.2 The Proposed STN Models 5.3 Network Visualisation 5.4 Comparing Three Targets with Increasing Difficulty 6 Discussion References GPAM: Genetic Programming with Associative Memory 1 Introduction 2 Related Work 2.1 Symbolic Regression and Genetic Programming 2.2 Efficient Processing of DNNs 2.3 Weight Compression 3 Proposed Method 3.1 The GPAM Approach 3.2 GPAM for Weight Generation 4 Results for Symbolic Regression Benchmarks 4.1 Benchmarks 4.2 Setup 4.3 Memory Sizing 4.4 Role of Constants in GPAM 5 Results for Weight Generation 6 Discussion and Conclusions References MAP-Elites with Cosine-Similarity for Evolutionary Ensemble Learning 1 Introduction 2 Related Work 2.1 Semantic GP 2.2 GP-Based Ensemble Learning 2.3 Quality Diversity Optimization 3 The Proposed Ensemble Learning Algorithm 3.1 The Overall Framework 3.2 Angle-Based Dimensionality Reduction 3.3 Reference Semantic Points 4 Experiment Settings 4.1 Datasets 4.2 Experimental Protocol 4.3 Parameter Settings 4.4 Benchmark Dimensionality Reduction Methods 5 Experimental Results 5.1 Comparisons of MAP-Elites Using Different Dimensionality Reduction Methods 5.2 Impact of Using Reference Points 5.3 Comparison with Other Machine Learning and Symbolic Regression Methods 6 Conclusions References Small Solutions for Real-World Symbolic Regression Using Denoising Autoencoder Genetic Programming 1 Introduction 2 Related Work 3 Denoising Autoencoder LSTM 3.1 Model Building and Sampling 3.2 A New Denoising Strategy: Levenshtein Tree Edit 4 Experiments 4.1 Experimental Setup 4.2 Prediction Quality 4.3 Analyzing the Search Behavior 5 Conclusions and Future Work References Context Matters: Adaptive Mutation for Grammars 1 Introduction 2 Background 2.1 Grammar-Based Genetic Programming 2.2 Adaptive Mutation Rate 2.3 Grammar-Design 3 Adaptive Facilitated Mutation 3.1 Grammar Design for Adaptive Facilitated Mutation 4 Experimental Setup 5 Results 6 Conclusion 6.1 Future Work References A Boosting Approach to Constructing an Ensemble Stack 1 Introduction 2 Related Work 3 Evolving an Ensemble Stack Using Boosting 3.1 The Boosting Ensemble Stack Algorithm 3.2 Evaluating an Ensemble Stack Post Training 3.3 Using an Extremely Large Number of Bins 4 Experimental Methodology 5 Results 5.1 Small Scale Classification Tasks 5.2 Large Scale Classification Task 6 Conclusion References Adaptive Batch Size CGP: Improving Accuracy and Runtime for CGP Logic Optimization Flow 1 Introduction 2 Cartesian Genetic Programming 2.1 Representation 2.2 Evolutionary Process 3 Methodology 3.1 Definitions 3.2 Adaptive Batch Size CGP 3.3 Experimental Protocol 4 Results 5 Conclusion References Faster Convergence with Lexicase Selection in Tree-Based Automated Machine Learning 1 Introduction 2 Related Work 3 Methods 3.1 Review of TPOT 3.2 Parent Selection Algorithms 4 Experimental Set-Up 4.1 Datasets 4.2 Implementation 4.3 Evaluating Convergence 4.4 Exploration of Pipelines 5 Results 5.1 DIGEN Datasets 5.2 ANGES Datasets 6 Discussion References Using FPGA Devices to Accelerate Tree-Based Genetic Programming: A Preliminary Exploration with Recent Technologies 1 Introduction 2 Related Work 3 Accelerator Architecture 3.1 Program Memory 3.2 Program Compiler 3.3 Program Evaluator 4 Design of Experiments 4.1 Comparison Metrics 4.2 Primitive Sets 4.3 Program Generation 4.4 Fitness Cases 5 Results 6 Current Limitations and Potential Optimizations 6.1 Current Limitations 6.2 Potential Optimizations 7 Conclusion References Memetic Semantic Genetic Programming for Symbolic Regression 1 Introduction 2 Semantic GP 2.1 Library Building and Searching 3 Memetic Algorithms 4 Memetic Semantic for Symbolic Regression 4.1 Algorithm 4.2 Local Tree Improvement 5 Experimental Setup 6 Results 7 Related Work 8 Conclusion References Grammatical Evolution with Code2vec 1 Introduction 2 Background 2.1 Grammatical Evolution 2.2 Code2vec 3 Methods 3.1 ClusterBooster 3.2 ClusterSelection 4 Experiments 4.1 Benchmarks 4.2 Experimental Setup 4.3 Results 5 Conclusion References Short Presentations Domain-Aware Feature Learning with Grammar-Guided Genetic Programming 1 Introduction 2 Related Work 2.1 Genetic-Programming-Based Feature Learning 2.2 Domain-Aware Feature Learning and Aggregation Incorporation 3 Method 3.1 Domain Knowledge M3GP 3.2 Domain Knowledge and Aggregation M3GP 4 Evaluation 4.1 Datasets 4.2 Implementation Details 4.3 Experiment Details 5 Results 6 Conclusion References Genetic Improvement of LLVM Intermediate Representation 1 Introduction 2 Background 3 Mutating LLVM IR 3.1 Representation 3.2 LLVM IR define Functions 3.3 Mutable LLVM IR 3.4 Compiling C/C++ etc. to Generate LLVM IR 3.5 Selecting Which LLVM IR to Optimise 3.6 Deleting LLVM IR 4 Fitness Function 4.1 Test Cases for Google's OLC and Uber's H3: GB Post Codes 4.2 Counting Instructions with perf stat -e instructions -x, 4.3 Sandboxing to Prevent Running Mutations Causing Harm 4.4 Timeouts to Stop Poor Mutants Delaying Search 4.5 Limiting Output Size to Avoid Filling Disk or Exceeding Disk Quota 5 Hillclimbing Search 6 Results 7 Discussion 7.1 Types of Improvement Found 7.2 Discussion: Future Work, Co-evolution, Perf, Fitness Landscape 8 Conclusions References Spatial Genetic Programming 1 Introduction 2 Related Literature 3 Spatial Genetic Programming 3.1 The Cost Function 3.2 Outputs, Termination Conditions and Model Execution 3.3 Evolution of Models and the Genetic Operators 3.4 Conditional Return Statements 4 Experiments and Results 4.1 Case Study: Classic Control Problems 4.2 Case Study: Custom Toy Problems 4.3 Impact of a Spatial Crossover on the Evolution of Programs 5 Conclusion References All You Need is Sex for Diversity 1 Introduction 2 The PIMP Approach 3 Motivation 4 Methodology 4.1 Measures 4.2 Statistical Tests 5 Results 6 Discussion and Additional Remarks 7 Conclusion References On the Effects of Collaborators Selection and Aggregation in Cooperative Coevolution: An Experimental Analysis 1 Introduction and Related Works 2 A General Scheme for CC 3 Case Studies 3.1 Toy Problems 3.2 Symbolic Regression 3.3 Neuroevolution 4 Experimental Analysis 4.1 Toy Problems 4.2 Symbolic Regression 4.3 Neuroevolution 5 Concluding Remarks References To Bias or Not to Bias: Probabilistic Initialisation for Evolving Dispatching Rules 1 Introduction 2 Background 2.1 Unrelated Machines Environment 2.2 Designing Dispatching Rules with Genetic Programming 3 Probabilistic Individual Initialisation 4 Experimental Analysis 4.1 Benchmark Setup 4.2 Results 5 Analysis 5.1 Node Probabilities 5.2 Method Ranking 6 Conclusion References MTGP: Combining Metamorphic Testing and Genetic Programming 1 Introduction 2 Related Work 3 Methodology 3.1 Metamorphic Testing 3.2 Benchmark Problems and Metamorphic Relations 3.3 Incorporating Metamorphic Testing in GP 4 Experiments and Discussion 4.1 Experimental Setup 4.2 Results and Discussion 5 Conclusion References Interacting Robots in an Artificial Evolutionary Ecosystem 1 Introduction 2 Related Work 3 Method 3.1 Evolutionary Algorithm 3.2 The Robots 3.3 Genotype Encodings 4 Experimental Setup 5 Analysis 5.1 Fitness Analysis 5.2 Trait Analysis 5.3 Differences Between Offspring Spawn Location 6 Conclusion References Author Index