Benjamin Inden

An examination of different fitness and novelty based selection methods for the evolution of neural networks

It has been suggested recently that it is a reasonable abstraction of evolutionary processes to use evolutionary algorithms that select individuals based on the novelty of their behavior instead of their fitness. Here we study the performance of fitness- and novelty-based search on several neuroevolution tasks. We also propose several new algorithms that select both for fit and for novel individuals, but without weighting these two criteria directly against each other. We find that behavioral speciation, behavioral near neutral speciation, and behavioral novelty speciation perform best on most tasks. Pure novelty search, as well as a number of hybrid methods without speciation mechanism, do not perform well on most tasks. Using behavioral criteria for speciation often yields better results than using genetic criteria.

Evolving neural fields for problems with large input and output spaces

We have developed an extension of the NEAT neuroevolution method, called NEATfields, to solve problems with large input and output spaces. The NEATfields method is a multilevel neuroevolution method using externally specified design patterns. Its networks have three levels of architecture. The highest level is a NEAT-like network of neural fields. The intermediate level is a field of identical subnetworks, called field elements, with a two-dimensional topology. The lowest level is a NEAT-like subnetwork of neurons. The topology and connection weights of these networks are evolved with methods derived from the NEAT method. Evolution is provided with further design patterns to enable information flow between field elements, to dehomogenize neural fields, and to enable detection of local features. We show that the NEATfields method can solve a number of high dimensional pattern recognition and control problems, provide conceptual and empirical comparison with the state of the art HyperNEAT method, and evaluate the benefits of different design patterns.

NEATfields: evolution of neural fields

We have developed a novel extension of the NEAT neuroevolution method, termed NEATfields, to solve problems with large input and output spaces. NEATfields networks are layered into two-dimensional fields of identical or similar subnetworks with an arbitrary topology. The subnetworks are evolved with genetic operations similar to those used in the NEAT neuroevolution method. We show that information processing within the neural fields can be organized by providing suitable building blocks to evolution. NEATfields can solve a number of visual discrimination tasks and a newly introduced multiple pole balancing task.

Open-ended coevolution and the emergence of complex irreducible functional units in iterated number sequence games

We present three related number sequence games as simple models of coevolution and demonstrate that they produce escalating arms races and irreducible functional units of unbounded size. We argue that our results imply that the models also show unbounded evolutionary activity according to a previous formal definition. Furthermore, we examine the robustness of the coevolutionary dynamics under different parameter regimes. We propose number sequence games as benchmarks for coevolutionary algorithms, and make some suggestions on adjusting task difficulty and choosing selection methods in coevolutionary algorithms.

Evolution of multisensory integration in large neural fields

We show that by evolving neural fields it is possible to study the evolution of neural networks that perform multisensory integration of high dimensional input data. In particular, four simple tasks for the integration of visual and tactile input are introduced. Neural networks evolve that can use these senses in a cost-optimal way, enhance the accuracy of classifying noisy input images, or enhance spatial accuracy of perception. An evolved neural network is shown to display a kind of McGurk effect.

Exploiting inherent regularity in control of multilegged robot locomotion by evolving neural fields

The control of multilegged robots is challenging because of the large number of sensors and actuators involved. However, the regularity inherent to gait control can be taken into account to design controllers for multilegged robots. In this paper, we show that NEATfields, a method designed for the evolution of large neural networks, can exploit this regularity to evolve significantly better gaits than those evolved by the standard NEAT method. We also show how evolved networks can control a robot with a ball-like morphology to move on a rough terrain. The success in evolving large neural networks suggests that the NEATfields method is a promising tool for studying complex behaviors in robotics and artificial life.

Neural agents can evolve to reproduce sequences of arbitrary length

We demonstrate that neural agents can evolve behavioral sequences of arbitrary length. In our framework, agents in a two-dimensional arena have to find the secure one among two possible patches, and which of them is secure changes over time. Evolution of arbitrarily long behavioral sequences is achieved by extending the neuroevolution method NEAT with two techniques: Only newly evolved network structure is subject to mutations, and inputs to the neural network are provided in an incremental fashion during evolution. It is suggested that these techniques are transferable to other neuroevolution methods and domains, and constitute a step towards achieving open-ended evolution. Furthermore, it is argued that the proposed techniques are strongly simplified models of processes that to some degree occur naturally in systems with more flexible genetic architectures.

Effects of Several Bioinspired Methods on the Stability of Coevolutionary Complexification

We study conditions for sustained growth of complexity in an abstract model of parasitic coevolution. Previous research has found that complexification is hard to achieve if the evolution of the symbiont population is constrained by the hosts but the evolution of the hosts is unconstrained, or, more generally, if the task difficulty is much higher for the symbionts than for the hosts. Here we study whether three bio inspired methods known from previous research on achieving stability in coevolution (balancing, niching, and reduced resistance) can restore complexification in such situations. We find that reduced resistance, and to a lesser degree niching, are successful if applied together with truncation selection, but not if applied together with fitness proportional selection.