I am a highly cited researcher in evolutionary computation, especially genetic programming (GP). I am best known as the inventor of a type of GP called Cartesian Genetic Programming (CGP) and also for a form of evolutionary computation using physical materials called evolution-in-materio.
Evolutionary algorithms have proved their worth on various optimization problems over the course ... more Evolutionary algorithms have proved their worth on various optimization problems over the course of years. However, some techniques like genetic programming (GP) and Cartesian genetic programming (CGP) are not restricted only to optimization problems but can be also used in classiication tasks. In this paper, we consider mixed-type CGP (MT-CGP) and test it on a number of benchmark binary and multi-class problems. Following that, we introduce a new representation for our algorithm where each node also has an accompanying weight factor called the amplitude. Our results suggest that this version of CGP is more powerful and able to obtain higher accuracies when compared to the mixed-type CGP or the standard CGP. Finally, we introduce the L1 regularization into MT-CGP in order to facilitate even further feature reduction.
This workshop follows on from the successful workshops on self-organization in representations in... more This workshop follows on from the successful workshops on self-organization in representations in evolutionary algorithms, and scalable, evolvable, emergent developmental systems at previous GECCO conferences. This year's workshop is a unified workshop covering both closely related areas. Evolutionary algorithms (EAs) have been applied to an ever increasing variety of problem domains, for which they have achieved human competitive results on small evolutionary design problems. The application of EAs ...
NeuroEvolution is the application of Evolutionary Algorithms to the training of Artificial Neural... more NeuroEvolution is the application of Evolutionary Algorithms to the training of Artificial Neural Networks. Currently the vast majority of NeuroEvolutionary methods create homogeneous networks of user defined transfer functions. This is despite Neuro-Evolution being capable of creating heterogeneous networks where each neuron's transfer function is not chosen by the user, but selected or optimised during evolution. This paper demonstrates how Neuro-Evolution can be used to select or optimise each neuron's transfer function and empirically shows that doing so significantly aids training. This result is important as most NeuroEvolutionary methods are capable of creating heterogeneous networks using the methods described .
In Cantu Paz E and Foster Ja and Deb K and Davis L and Roy R and O Reilly U M and Beyer H G and Standish Rk and Kendall G and Wilson Sw and Harman M and Wegener J and Dasgupta D and Potter Ma and Schultz Ac and Dowsland Ka and Jonoska N and Miller Jf Gecco Springer, 2003
Modern FPGAs provide a platform for implementation of uncommitted logic a rrays which are a lso, ... more Modern FPGAs provide a platform for implementation of uncommitted logic a rrays which are a lso, in many cases, reconfigurable. Whilst t his allows circuit functionality to b e c hanged in time, it also p rovides a c onvenient environment i n which to encourage the direct evolution (using g enetic a lgorithms) of those c ircuit solutions themselves.
Genetic and Evolutionary Computation Conference, 2000
We argue that there is an upper limit on the complexity of software that can be constructed using... more We argue that there is an upper limit on the complexity of software that can be constructed using current methods. Furthermore, this limit is orders of mag- nitude smaller than the complexity of living systems. We argue that many of the ad- vantages of autonomic computing will not be possible unless fundamental aspects of living systems are incorporated into a
Evolutionary algorithms have proved their worth on various optimization problems over the course ... more Evolutionary algorithms have proved their worth on various optimization problems over the course of years. However, some techniques like genetic programming (GP) and Cartesian genetic programming (CGP) are not restricted only to optimization problems but can be also used in classiication tasks. In this paper, we consider mixed-type CGP (MT-CGP) and test it on a number of benchmark binary and multi-class problems. Following that, we introduce a new representation for our algorithm where each node also has an accompanying weight factor called the amplitude. Our results suggest that this version of CGP is more powerful and able to obtain higher accuracies when compared to the mixed-type CGP or the standard CGP. Finally, we introduce the L1 regularization into MT-CGP in order to facilitate even further feature reduction.
This workshop follows on from the successful workshops on self-organization in representations in... more This workshop follows on from the successful workshops on self-organization in representations in evolutionary algorithms, and scalable, evolvable, emergent developmental systems at previous GECCO conferences. This year's workshop is a unified workshop covering both closely related areas. Evolutionary algorithms (EAs) have been applied to an ever increasing variety of problem domains, for which they have achieved human competitive results on small evolutionary design problems. The application of EAs ...
NeuroEvolution is the application of Evolutionary Algorithms to the training of Artificial Neural... more NeuroEvolution is the application of Evolutionary Algorithms to the training of Artificial Neural Networks. Currently the vast majority of NeuroEvolutionary methods create homogeneous networks of user defined transfer functions. This is despite Neuro-Evolution being capable of creating heterogeneous networks where each neuron's transfer function is not chosen by the user, but selected or optimised during evolution. This paper demonstrates how Neuro-Evolution can be used to select or optimise each neuron's transfer function and empirically shows that doing so significantly aids training. This result is important as most NeuroEvolutionary methods are capable of creating heterogeneous networks using the methods described .
In Cantu Paz E and Foster Ja and Deb K and Davis L and Roy R and O Reilly U M and Beyer H G and Standish Rk and Kendall G and Wilson Sw and Harman M and Wegener J and Dasgupta D and Potter Ma and Schultz Ac and Dowsland Ka and Jonoska N and Miller Jf Gecco Springer, 2003
Modern FPGAs provide a platform for implementation of uncommitted logic a rrays which are a lso, ... more Modern FPGAs provide a platform for implementation of uncommitted logic a rrays which are a lso, in many cases, reconfigurable. Whilst t his allows circuit functionality to b e c hanged in time, it also p rovides a c onvenient environment i n which to encourage the direct evolution (using g enetic a lgorithms) of those c ircuit solutions themselves.
Genetic and Evolutionary Computation Conference, 2000
We argue that there is an upper limit on the complexity of software that can be constructed using... more We argue that there is an upper limit on the complexity of software that can be constructed using current methods. Furthermore, this limit is orders of mag- nitude smaller than the complexity of living systems. We argue that many of the ad- vantages of autonomic computing will not be possible unless fundamental aspects of living systems are incorporated into a
Uploads
Papers by Julian Miller