Wolfgang Banzhaf

Wolfgang Banzhaf
Michigan State University | MSU · Department of Computer Science and Engineering

PhD in Physics from Fridericiana (now KIT) Karlsruhe

About

445
Publications
47,290
Reads
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
13,308
Citations
Introduction
I am a researcher in bio-inspired computing. My interests range from principles of life and evolution, via evolutionary algorithms (in particular genetic programming) to applications in image processing (deep neural networks for image processing and other tasks) and explainable model generation using GP. I am also involved in research on social, economic and ecological questions.
Additional affiliations
September 2016 - present
Michigan State University
Position
  • Chair
September 2003 - July 2023
Memorial University of Newfoundland
Position
  • Professor
October 1993 - September 2003
Technische Universität Dortmund
Position
  • Universität Dortmund
Education
September 1982 - July 1985
Karlsruhe Institute of Technology
Field of study
  • Theoretical Physics

Publications

Publications (445)
Preprint
Full-text available
The use of correlation as a fitness function is explored in symbolic regression tasks and the performance is compared against the typical RMSE fitness function. Using correlation with an alignment step to conclude the evolution led to significant performance gains over RMSE as a fitness function. Using correlation as a fitness function led to solut...
Conference Paper
Full-text available
Genetic improvement is a search technique that aims to improve a given acceptable solution to a problem. In this paper, we present the novel use of genetic improvement to find problem-specific optimized LLVM pass sequences. We develop a pass-level patch representation in the linear genetic programming framework, Shackleton, to evolve the modificati...
Preprint
Full-text available
Genetic improvement is a search technique that aims to improve a given acceptable solution to a problem. In this paper, we present the novel use of genetic improvement to find problem-specific optimized LLVM pass sequences. We develop a pass-level patch representation in the linear genetic programming framework, Shackleton, to evolve the modificati...
Preprint
Proteins are used by scientists to serve a variety of purposes in clinical practice and laboratory research. To optimize proteins for greater function, a variety of techniques have been developed. For the development of reporter genes used in Magnetic Resonance Imaging (MRI) based on Chemical Exchange Saturation Transfer (CEST), these techniques ha...
Preprint
Program synthesis aims to {\it automatically} find programs from an underlying programming language that satisfy a given specification. While this has the potential to revolutionize computing, how to search over the vast space of programs efficiently is an unsolved challenge in program synthesis. In cases where large programs are required for a sol...
Preprint
Full-text available
In this paper we introduce an active learning method for symbolic regression using StackGP. The approach begins with a small number of data points for StackGP to model. To improve the model the system incrementally adds a data point such that the new point maximizes prediction uncertainty as measured by the model ensemble. Symbolic regression is re...
Preprint
Full-text available
In this paper we introduce Shackleton as a generalized framework enabling the application of linear genetic programming -- a technique under the umbrella of evolutionary algorithms -- to a variety of use cases. We also explore here a novel application for this class of methods: optimizing sequences of LLVM optimization passes. The algorithm underpi...
Preprint
The Genetically Evolved NEutrino Telescopes for Improved Sensitivity, or GENETIS, project seeks to optimize detectors in physics for science outcomes in high dimensional parameter spaces. In this project, we designed an antenna using a genetic algorithm with a science outcome directly as the sole figure of merit. This paper presents initial results...
Article
Full-text available
A fundamental aspect of intelligent agent behaviour is the ability to encode salient features of experience in memory and use these memories, in combination with current sensory information, to predict the best action for each situation such that long-term objectives are maximized. The world is highly dynamic, and behavioural agents must generalize...
Article
Modularity represents a recurring theme in the attempt to scale evolution to the design of complex systems. However, modularity rarely forms the central theme of an artificial approach to evolution. In this work, we report on progress with the recently proposed Tangled Program Graph (TPG) framework in which programs are modules. The combination of...
Article
Full-text available
We consider a number of Artificial Chemistry models for economic activity and what consequences they have for the formation of economic inequality. We are particularly interested in what tax measures are effective in dampening economic inequality. By starting from well-known kinetic exchange models, we examine different scenarios for reducing the t...
Poster
Full-text available
The video game Factorio by Wube Software*, has a myriad of gameplay scenarios which are analogous to real-world problems, and is a useful simulator for developing solutions for them. In this paper we define the logistic transport belt problem, we provide an interface to allow optimizers in any programming language to interact with Factorio, and we...
Poster
Full-text available
LLVM IR (low-level virtual machine intermediate representation) is an intermediate step in the compilation of computer code. LLVM compilers allow optimization by using a sequence of steps (passes) to improve run-time or other criteria. Genetic Programming (GP) is an algorithm that is inspired by the natural selection process and can automatically g...
Preprint
Full-text available
A fundamental aspect of behaviour is the ability to encode salient features of experience in memory and use these memories, in combination with current sensory information, to predict the best action for each situation such that long-term objectives are maximized. The world is highly dynamic, and behavioural agents must generalize across a variety...
Preprint
Full-text available
Efficient optimization of resources is paramount to success in many problems faced today. In the field of operational research the efficient scheduling of employees; packing of vans; routing of vehicles; logistics of airlines and transport of materials can be the difference between emission reduction or excess, profits or losses and feasibility or...
Article
Neural architecture search (NAS) has emerged as a promising avenue for automatically designing task-specific neural network architectures. Most existing NAS approaches require one complete search for each deployment specification of hardware or objective. This is a computationally impractical endeavor given the potentially large number of applicati...
Chapter
In this paper, we propose an efficient NAS algorithm for generating task-specific models that are competitive under multiple competing objectives. It comprises of two surrogates, one at the architecture level to improve sample efficiency and one at the weights level, through a supernet, to improve gradient descent training efficiency. On standard b...
Article
Convolutional neural networks (CNNs) are the backbones of deep learning paradigms for numerous vision tasks. Early advancements in CNN architectures are primarily driven by human expertise and by elaborate design processes. Recently, neural architecture search was proposed with the aim of automating the network design process and generating task-de...
Article
Full-text available
Genotype–phenotype mapping plays an essential role in the design of an evolutionary algorithm. Variation occurs at the genotypic level but fitness is evaluated at the phenotypic level, therefore, this mapping determines if and how variations are effectively translated into quality improvements. In evolutionary algorithms, this mapping has often bee...
Preprint
In this paper, we propose an efficient NAS algorithm for generating task-specific models that are competitive under multiple competing objectives. It comprises of two surrogates, one at the architecture level to improve sample efficiency and one at the weights level, through a supernet, to improve gradient descent training efficiency. On standard b...
Chapter
The use of genetic variation (DNA markers) has become widespread for prediction of genetic merit in animal and plant breeding and it is gaining momentum as a prognostic tool for propensity to disease in human medicine. Although conceptually straightforward, genomic prediction is a very challenging problem. Genotyping organisms and recording phenoty...
Chapter
The automation of program repair can be coached in terms of search algorithms. Repair templates derived from common bug-fix patterns can be used to determine a promising search space with potentially many correct patches, a space that can be effectively explored by GP methods. Here we propose a new repair system, ARJA-p, extended from our earlier A...
Preprint
We consider a number of Artificial Chemistry models for economic activity and what consequences they have for the formation of economic inequality. We are particularly interested in what tax measures are effective in dampening economic inequality. By starting from well-known kinetic exchange models, we examine different scenarios for reducing the t...
Conference Paper
Convolutional neural networks (CNNs) are the backbones of deep learning paradigms for numerous vision tasks. Early advancements in CNN architectures are primarily driven by human expertise and elaborate design. Recently, neural architecture search (NAS) was proposed with the aim of automating the network design process and generating task-dependent...
Preprint
Neural architecture search (NAS) has emerged as a promising avenue for automatically designing task-specific neural networks. Most existing NAS approaches require one complete search for each deployment specification of hardware or objective. This is a computationally impractical endeavor given the potentially large number of application scenarios....
Chapter
Video games provide a well-defined study ground for the development of behavioural agents that learn through trial-and-error interaction with their environment, or reinforcement learning (RL). They cover a diverse range of environments that are designed to be challenging for humans, all through a high-dimensional visual interface. Tangled Program G...
Chapter
The present and future of evolutionary algorithms depends on the proper use of modern parallel and distributed computing infrastructures. Although still sequential approaches dominate the landscape, available multi-core, many-core and distributed systems will make users and researchers to more frequently deploy parallel version of the algorithms. I...
Preprint
Full-text available
The present and future of evolutionary algorithms depends on the proper use of modern parallel and distributed computing infrastructures. Although still sequential approaches dominate the landscape, available multi-core, many-core and distributed systems will make users and researchers to more frequently deploy parallel version of the algorithms. I...
Chapter
When a test suite is considered as the specification, the paradigm is called test-suite based repair Monperrus (ACM Comput Surv 51(1):17, 2018). The test suite should contain at least one negative (i.e., initially failing) test that triggers the bug to be fixed and a number of positive (i.e., initially passing) tests that define the expected progra...
Chapter
This chapter proposes a new model of tree-based Genetic Programming (GP) as a simple sampling algorithm that samples minimal schemata (subsets of the solution space) described by a single concrete node at a single position in the expression tree. We show that GP explores these schemata in the same way across three benchmarks, rapidly converging the...
Article
Full-text available
This paper presents a new way of measuring complexity in variable-size-chromosome-based evolutionary algorithms. Dealing with complexity is particularly useful when considering bloat in Genetic Programming. Instead of analyzing size growth, we focus on the time required for individuals’ fitness evaluations, which correlates with size. This way, we...
Chapter
Full-text available
This contribution argues that the notion of time used in the scientific modeling of reality deprives time of its real nature. Difficulties from logic paradoxes to mathematical incompleteness and numerical uncertainty ensue. How can the emergence of novelty in the Universe be explained? How can the creativity of the evolutionary process leading to e...
Book
This edited research monograph brings together contributions from computer scientists, biologists, and engineers who are engaged with the study of evolution and how it may be applied to solve real-world problems. It also serves as a Festschrift dedicated to Erik D. Goodman, the founding director of the BEACON Center for the Study of Evolution in Ac...
Book
These contributions, written by the foremost international researchers and practitioners of Genetic Programming (GP), explore the synergy between theoretical and empirical results on real-world problems, producing a comprehensive view of the state of the art in GP. In this year’s edition, the topics covered include many of the most important issues...
Article
Bug repair is a major component of software maintenance, which requires a huge amount of manpower. Evolutionary computation, particularly genetic programming (GP), is a class of promising techniques for automating this time-consuming and expensive process. Although recent research in evolutionary program repair has made significant progress, major...
Preprint
Convolutional neural networks (CNNs) are the backbones of deep learning paradigms for numerous vision tasks. Early advancements in CNN architectures are primarily driven by human expertise and elaborate design. Recently, neural architecture search was proposed with the aim of automating the network design process and generating task-dependent archi...
Chapter
Evolutionary computation (EC) is the area of computer science and engineering that concerns itself with algorithms derived from formalizing natural evolution. This is part of a larger effort to draw inspiration from biological systems for computational purposes. Evolutionary computation methods have been used to solve optimization problems, to mode...
Conference Paper
We evolve floating point Sextic polynomial populations of genetic programming binary trees for up to a million generations. Programs with almost 400 000 000 instructions are created by crossover. To support unbounded Long-Term Evolution Experiment LTEE GP we use both SIMD parallel AVX 512 bit instructions and 48 threads to yield performance of up t...
Conference Paper
This paper presents an automatic software repair system that combines the characteristic components of several typical evolutionary computation based repair approaches into a unified repair framework so as to take advantage of their respective component strengths. We exploit both the redundancy assumption and repair templates to create a search spa...
Conference Paper
Full-text available
Lexicase selection achieves very good solution quality by introducing ordered test cases. However, the computational complexity of lexicase selection can prohibit its use in many applications. In this paper, we introduce Batch Tournament Selection (BTS), a hybrid of tournament and lexicase selection which is approximately one order of magnitude fas...
Conference Paper
This paper introduces NSGA-Net --- an evolutionary approach for neural architecture search (NAS). NSGA-Net is designed with three goals in mind: (1) a procedure considering multiple and conflicting objectives, (2) an efficient procedure balancing exploration and exploitation of the space of potential neural network architectures, and (3) a procedur...
Conference Paper
This paper presents a new multi-gene genetic programming (MGGP) approach for prediction of bond strength of Fiber Reinforced Polymers (FRP) bars to concrete. The MGGP technique models the bond behavior of FRP bars by integrating the capabilities of standard genetic programming and classical regression. The factors that affect the bond strength of F...
Article
Full-text available
We have shown previously that a feed-forward, back propagation neural network model based on composite n-grams can predict normalized signal strengths of a microarray based DNA sequencing experiment. The microarray comprises a 4xN set of 25-base single-stranded DNA molecule ('oligos'), specific for each of the four possible bases (A, C, G, or T) fo...
Article
Full-text available
Accurate prediction of bond behavior of fiber reinforcement polymer (FRP) concrete has a pivotal role in the construction industry. This paper presents a soft computing method called multi-gene genetic programming (MGGP) to develop an intelligent prediction model for the bond strength of FRP bars in concrete. The main advantage of the MGGP method o...
Conference Paper
Full-text available
The present and future of evolutionary algorithms depends on the proper use of modern parallel and distributed computing infrastructures. Although still sequential approaches dominate the landscape, available multi-core, many-core and distributed systems will make users and researchers to more frequently deploy parallel version of the algorithms. I...
Preprint
Lexicase selection achieves very good solution quality by introducing ordered test cases. However, the computational complexity of lexicase selection can prohibit its use in many applications. In this paper, we introduce Batch Tournament Selection (BTS), a hybrid of tournament and lexicase selection which is approximately one order of magnitude fas...
Chapter
The genotype-to-phenotype mapping plays an essential role in the design of an evolutionary algorithm. Since variation occurs at the genotypic level but fitness is evaluated at the phenotypic level, this mapping determines how variations are effectively translated into quality improvements. We numerically study the redundant genotype-to-phenotype ma...
Preprint
We evolve floating point Sextic polynomial populations of genetic programming binary trees for up to a million generations. Programs with almost four hundred million instructions are created by crossover. To support unbounded Long-Term Evolution Experiment LTEE GP we use both SIMD parallel AVX 512 bit instructions and 48 threads to yield performanc...
Article
Full-text available
In nature, gene regulatory networks are a key mediator between the information stored in the DNA of living organisms (their genotype) and the structural and behavioral expression this finds in their bodies, surviving in the world (their phenotype). They integrate environmental signals, steer development, buffer stochasticity, and allow evolution to...
Preprint
This contribution argues that the notion of time used in the scientific modeling of reality deprives time of its real nature. Difficulties from logic paradoxes to mathematical incompleteness and numerical uncertainty ensue. How can the emergence of novelty in the Universe be explained? How can the creativity of the evolutionary process leading to e...
Book
These contributions, written by the foremost international researchers and practitioners of Genetic Programming (GP), explore the synergy between theoretical and empirical results on real-world problems, producing a comprehensive view of the state of the art in GP. Topics in this volume include: evolving developmental programs for neural networks s...
Preprint
Full-text available
Evolutionary algorithms often incorporate ecological concepts to help maintain diverse populations and drive continued innovation. However, while there is strong evidence for the value of ecological dynamics, a lack of overarching theoretical framework renders the precise mechanisms behind these results unclear. These gaps in our understanding make...
Preprint
Evolutionary algorithms often incorporate ecological concepts to help maintain diverse populations and drive continued innovation. However, while there is strong evidence for the value of ecological dynamics, a lack of overarching theoretical framework renders the precise mechanisms behind these results unclear. These gaps in our understanding make...
Article
Full-text available
This paper proposes Drone Squadron Optimization (DSO), a new self-adaptive metaheuristic for global numerical optimization which is updated online by a hyper-heuristic. DSO is an artifact-inspired technique, as opposed to many nature-inspired algorithms used today. DSO is very flexible because it is not related to natural behaviors or phenomena. DS...
Preprint
This paper introduces NSGA-Net, an evolutionary approach for neural architecture search (NAS). NSGA-Net is designed with three goals in mind: (1) a NAS procedure for multiple, possibly conflicting, objectives, (2) efficient exploration and exploitation of the space of potential neural network architectures, and (3) output of a diverse set of networ...
Conference Paper
We present AutoMap, a pair of methods for automatic generation of evolvable genotype-phenotype mappings. Both use an artificial neural network autoencoder trained on phenotypes harvested from fitness peaks as the basis for a genotype-phenotype mapping. In the first, the decoder segment of a bottlenecked autoencoder serves as the genotype-phenotype...
Chapter
In this chapter we take a fresh look at the current status of evolving computer code using Genetic Programming methods. The emphasis is not so much on what has been achieved in detail in the past few years, but on the general research direction of code evolution and its ramifications for GP. We begin with a quick glance at the area of Search-based...
Chapter
Redundant mapping from genotype to phenotype is common in evolutionary algorithms, especially in genetic programming (GP). Such a redundancy leads to neutrality, a situation where mutations to a genotype may not alter its phenotypic outcome. The effects of neutrality can be better understood by quantitatively analyzing its two observed properties,...
Article
Full-text available
Recent empirical studies show that the performance of GenProg is not satisfactory, particularly for Java. In this paper, we propose ARJA, a new GP based repair approach for automated repair of Java programs. To be specific, we present a novel lower-granularity patch representation that properly decouples the search subspaces of likely-buggy locatio...
Article
Full-text available
Symbolic Regression (SR) is a well-studied task in Evolutionary Computation (EC), where adequate free-form mathematical models must be automatically discovered from observed data. Statisticians, engineers, and general data scientists still prefer traditional regression methods over EC methods because of the solid mathematical foundations, the inter...
Article
Full-text available
A microarray DNA sequencing experiment for a molecule of N bases produces a 4xN data matrix, where for each of the N positions each quartet comprises the signal strength of binding of an experimental DNA to a reference oligonucleotide affixed to the microarray, for the four possible bases (A, C, G, or T). The strongest signal in each quartet should...
Conference Paper
In this contribution we propose a hybrid genetic programming approach for evolving a decision making system in the domain of RoboCup Soccer (Simulation League). Genetic programming has been rarely used in this domain in the past, due to the difficulties and restrictions of the soccer simulation. The real-time requirements of robot soccer and the le...
Article
Predicting the properties of materials like concrete has been proven a difficult task given the complex interactions among its components. Over the years, researchers have used Statistics, Machine Learning, and Evolutionary Computation to build models in an attempt to accurately predict such properties. High-quality models are often non-linear, jus...
Chapter
A data set for classification is commonly composed of a set of features defining the data space representation and one attribute corresponding to the instances’ class. A classification tool has to discover how to separate classes based on features, but the discovery of useful knowledge may be hampered by inadequate or insufficient features. Pre-pro...
Chapter
Feature selecting is considered as one of the most important pre-process methods in machine learning, data mining and bioinformatics. By applying pre-process techniques, we can defy the curse of dimensionality by reducing computational and storage costs, facilitate data understanding and visualization, and diminish training and testing times, leadi...