Search

It is already well known that environmental variation has a big effect on real evolution, and similar effects have been found in evolutionary artificial life simulations. In particular, a lot of research has been carried out on how the various evolutionary outcomes depend on the noise distributions representing the environmental changes, and how important it is for models to use inverse power-law distributions with the right noise colour. However, there are two distinct factors of relevance—the average total magnitude of change per unit time and the distribution of individual change magnitudes—and misleading results may emerge if those factors are not properly separated. This article makes use of an existing agent-based artificial life modeling framework to explore this issue using models previously tried and tested for other purposes. It begins by demonstrating how the total magnitude and distribution effects can easily be confused, and goes on to show how it is possible to untangle the influence of these interacting factors by using correlation-based normalization. It then presents a series of simulation results demonstrating that interesting dependencies on the noise distribution remain after separating those factors, but many effects involving the noise colour of inverse power-law distributions disappear, and very similar results arise across restricted-range white-noise distributions. The average total magnitude of change per unit time is found to have a substantial effect on the simulation outcomes, but the distribution of individual changes has very little effect. A robust counterexample is thereby provided to the idea that it is always important to use accurate environmental change distributions in artificial life models.

Learning is an important aspect of cognition that is crucial for the success of many species, and has been a factor involved in the evolution of distinct patterns of life history that depend on the environments in question. The extent to which different degrees of social and individual learning emerge follows from various species-dependent factors, such as the fidelity of information transmission between individuals, and that has previously been modelled in agent-based simulations with meme-based representations of learned knowledge and behaviours. A limitation of that previous work is that it was based on fixed environments, and it is known that different learning strategies will emerge depending on the variability of the environment. This paper will address that limitation by extending the existing modelling framework to allow the simulation of life history evolution and the emergence of appropriate learning strategies in changing environments.

The idea that lifetime learning can have a significant effect on life history evolution has recently been explored using a series of artificial life simulations. These involved populations of competing individuals evolving by natural selection to learn to perform well on simplified abstract tasks, with the learning consisting of identifying regularities in their environment. In reality, there is more to learning than that type of direct individual experience, because it often includes a substantial degree of social learning that involves various forms of imitation of what other individuals have learned before them. This article rectifies that omission by incorporating memes and imitative learning into revised versions of the previous approach. To do this reliably requires formulating and testing a general framework for meme-based simulations that will enable more complete investigations of learning as a factor in any life history evolution scenarios. It does that by simulating imitative information transfer in terms of memes being passed between individuals, and developing a process for merging that information with the (possibly inconsistent) information acquired by direct experience, leading to a consistent overall body of learning. The proposed framework is tested on a range of learning variations and a representative set of life history factors to confirm the robustness of the approach. The simulations presented illustrate the types of interactions and tradeoffs that can emerge, and indicate the kinds of species-specific models that could be developed with this approach in the future.

Display Omitted A multiobjective evolutionary algorithm for vehicle routing problems with backhauls.Our approach equals or improves upon some of the best-known single-objective results.Instances have few bi-objective conflicts; SSMOEA, NSGA-II and MOEA/D perform well.Large tri-objective solution sets and significant differences in both algorithms.Triobjective setting finds comparable solutions to those from bi-objective settings. The vehicle routing problem (VRP) is an important aspect of transportation logistics with many variants. This paper studies the VRP with backhauls (VRPB) in which the set of customers is partitioned into two subsets: linehaul customers requiring a quantity of product to be delivered, and backhaul customers with a quantity to be picked up. The basic VRPB involves finding a collection of routes with minimum cost, such that all linehaul and backhaul customers are serviced. A common variant is the VRP with selective backhauls (VRPSB), where the collection from backhaul customers is optional. For most real world applications, the number of vehicles, the total travel cost, and the uncollected backhauls are all important objectives to be minimized, so the VRPB needs to be tackled as a multi-objective problem. In this paper, a similarity-based selection evolutionary algorithm approach is proposed for finding improved multi-objective solutions for VRPB, VRPSB, and two further generalizations of them, with fully multi-objective performance evaluation.

An agent-based simulation framework is presented that provides a principled approach for investigating gender inequalities in professional hierarchies such as universities or businesses. Populations of artificial agents compete for promotion in their chosen professions, leading to emergent distributions that can be matched to real-life scenarios, and allowing the influence of socially or genetically acquired career preferences to be explored. The aim is that such models will enable better understanding of how imbalances emerge and evolve, facilitate the identification of specific signals that can indicate the presence or absence of discrimination, and provide a tool for determining how and when particular intervention strategies may be appropriate for rectifying any inequalities. Results generated from a representative series of abstract case studies involving innate or culturally-acquired gender-based ability differences, gender-based discrimination, and various forms of gender-specific career preferences, demonstrate the power of the approach. These simulations will hopefully inspire and facilitate better approaches for dealing with these issues in real life.

The Artificial Bee Colony (ABC) is a swarm intelligence algorithm for optimization that has previously been applied to the training of neural networks. This paper examines more carefully the performance of the ABC algo- rithm for optimizing the connection weights of feed-forward neural networks for classification tasks, and presents a more rigorous comparison with the traditional Back-Propagation (BP) training algorithm. The empirical results for bench- mark problems demonstrate that using the standard "stop- ping early" approach with optimized learning parameters leads to improved BP performance over the previous com- parative study, and that a simple variation of the ABC approach provides improved ABC performance too. With bothimprovementsapplied,theABCapproachdoesperform very well on small problems, but the generalization perfor- mances achieved are only significantly better than standard BP on one out of six datasets, and the training times increase rapidly as the size of the problem grows. If different, evo- lutionary optimized, BP learning rates are allowed for the two layers of the neural network, BP is significantly better than the ABC on two of the six datasets, and not significantly different on the other four.

In a previous article, we presented a systematic computational study of the extraction of semantic representations from the word-word co-occurrence statistics of large text corpora. The conclusion was that semantic vectors of pointwise mutual information values from very small co-occurrence windows, together with a cosine distance measure, consistently resulted in the best representations across a range of psychologically relevant semantic tasks. This article extends that study by investigating the use of three further factors--namely, the application of stop-lists, word stemming, and dimensionality reduction using singular value decomposition (SVD)--that have been used to provide improved performance elsewhere. It also introduces an additional semantic task and explores the advantages of using a much larger corpus. This leads to the discovery and analysis of improved SVD-based methods for generating semantic representations (that provide new state-of-the-art performance on a standard TOEFL task) and the identification and discussion of problems and misleading results that can arise without a full systematic study.

The vehicle routing problem with time windows is a complex combinatorial problem with many real-world applications in transportation and distribution logistics. Its main objective is to find the lowest distance set of routes to deliver goods, using a fleet of identical vehicles with restricted capacity, to customers with service time windows. However, there are other objectives, and having a range of solutions representing the trade-offs between objectives is crucial for many applications. Although previous research has used evolutionary methods for solving this problem, it has rarely concentrated on the optimization of more than one objective, and hardly ever explicitly considered the diversity of solutions. This paper proposes and analyzes a novel multi-objective evolutionary algorithm, which incorporates methods for measuring the similarity of solutions, to solve the multi-objective problem. The algorithm is applied to a standard benchmark problem set, showing that when the similarity measure is used appropriately, the diversity and quality of solutions is higher than when it is not used, and the algorithm achieves highly competitive results compared with previously published studies and those from a popular evolutionary multi-objective optimizer.

An artificial life approach is taken to explore the effect that lifetime learning can have on the evolution of certain life history traits, in particular the periods of protection that parents offer their young, and the age at first reproduction of those young. The study begins by simulating the evolution of simple artificial neural network systems that must learn quickly to perform well on simple classification tasks, and determining if and when extended periods of parental protection emerge. It is concluded that longer periods of parental protection of children do offer clear learning advantages and better adult performance, but only if procreation is not allowed during the protection period. In this case, a compromise protection period evolves that balances the improved learning performance against reduced procreation period. The crucial properties of the neural learning processes are then abstracted out to explore the possibility of studying the effect of learning more generally and with better computational efficiency. Throughout, the implications of these simulations for more realistic scenarios are discussed.

This paper presents two novel genetic algorithms (GAs) for hard industrially relevant packing problems. The design of both algorithms is inspired by aspects of molecular genetics, in particular, the modular exon-intron structure of eukaryotic genes. Two representative packing problems are used to test the utility of the proposed approach: the bin packing problem (BPP) and the multiple knapsack problem (MKP). The algorithm for the BPP, the exon shuffling GA (ESGA), is a steady-state GA with a sophisticated crossover operator that makes maximum use of the principle of natural selection to evolve feasible solutions with no explicit verification of constraint violations. The second algorithm, the Exonic GA (ExGA), implements an RNA inspired adaptive repair function necessary for the highly constrained MKP. Three different variants of this algorithm are presented and compared, which evolve a partial ordering of items using a segmented encoding that is utilised in the repair of infeasible solutions. All algorithms are tested on a range of benchmark problems, and the results indicate a very high degree of accuracy and reliability compared to other approaches in the literature.

The idea that at least some aspects of word meaning can be induced from patterns of word co-occurrence is becoming increasingly popular. However, there is less agreement about the precise computations involved, and the appropriate tests to distinguish between the various possibilities. It is important that the effect of the relevant design choices and parameter values are understood if psychological models using these methods are to be reliably evaluated and compared. In this article, we present a systematic exploration of the principal computational possibilities for formulating and validating representations of word meanings from word co-occurrence statistics. We find that, once we have identified the best procedures, a very simple approach is surprisingly successful and robust over a range of psychologically relevant evaluation measures.

Modularity in the human brain remains a controversial issue, with disagreement over the nature of the modules that exist, and why, when, and how they emerge. It is a natural assumption that modularity offers some form of computational advantage, and hence evolution by natural selection has translated those advantages into the kind of modular neural structures familiar to cognitive scientists. However, simulations of the evolution of simplified neural systems have shown that, in many cases, it is actually non-modular architectures that are most efficient. In this paper, the relevant issues are discussed and a series of simulations are presented that reveal crucial dependencies on the details of the learning algorithms and tasks that are being modelled, and the importance of taking into account known physical brain constraints, such as the degree of neural connectivity. A pattern is established which provides one explanation of why modularity should emerge reliably across a range of neural processing tasks.

PurposeThe purpose of this paper is to discuss the application of computational intelligence techniques to the field of industrial robot control.Design/methodology/approachThe core ideas behind using neural computation, evolutionary computation, and fuzzy logic techniques are presented, along with a selection of specific real‐world applications.FindingsTheir practical advantages and disadvantages relative to more traditional approaches are made clear.Originality/valueThe reader will appreciate the power of computational intelligence techniques for industrial robot control, and hopefully be encouraged to explore further the possibility of using them to achieve improved performance in their own applications.

Autonomous neural network systems typically require fast learning and good generalization performance, and there is potentially a trade-off between the two. The use of evolutionary techniques to improve the learning abilities of neural network systems is now widespread. However, there are a range of different evolutionary approaches that could be applied, and no systematic investigation has been carried out to find which work best. In this paper, such an investigation is presented, and it is shown that a range of evolutionary techniques can generate high performance networks, but they often lead to unwanted side effects, such as occasional instances of very poor performance. The nature of these problems are explored further, and it is shown how the evolution of age dependent plasticities and/or the use of ensemble techniques can alleviate them. A range of techniques are thus identified, with differing properties, that can be matched to the specific requirements of each application.

Attempts to formulate realistic models of the development of the human oculomotor control system have led to the conclusion that evolutionary factors play a crucial role. Moreover, even rather coarse simulations of the biological evolutionary processes result in adaptable control systems that are considerably more efficient than those designed by human researchers. In this paper I shall describe some of the aspects of these biological models that are likely to be useful for building robot control systems. In particular, I shall consider the evolution of appropriate innate starting points for learning/adaptation, patterns of learning rates that vary across different system components, learning rates that vary during the system's lifetime, and the relevance of individual differences across the evolved populations.

Setting up a neural network with a learning algorithm that determines how it can best operate is an efficient way to formulate control systems for many engineering applications, and is often much more feasible than direct programming. This paper examines three important aspects of this approach: the details of the cost function that is used with the gradient descent learning algorithm, how the resulting system depends on the initial pre-learning connection weights, and how the resulting system depends on the pattern of learning rates chosen for the different components of the system. We explore these issues by explicit simulations of a toy model that is a simplified abstraction of part of the human oculomotor control system. This allows us to compare our system with that produced by human evolution and development. We can then go on to consider how we might improve on the human system and apply what we have learnt to control systems that have no human analogue.

Recent work has demonstrated that counts of which other words co-occur with a word of interest can reflect interesting properties of that word. We have studied aspects of this kind of methodology by systematically examining the effects of different combinations of parameters used in the preparation of co-occurrence statistics. Several psychologically relevant evaluation measures are used. We have found that successful performance on the evaluation tasks depends on the correct selection of parameters such as window size and distance metric.

The Artificial Bee Colony (ABC) is a recently introduced swarm intelligence algorithm for optimization, that has previously been applied successfully to the training of neural networks. This paper explores more carefully the performance of the ABC algorithm for optimizing the connection weights of feed-forward neural networks for classification tasks, and presents a more rigorous comparison with the traditional Back-Propagation (BP) training algorithm. The empirical results show that using the standard “stopping early” approach with optimized learning parameters leads to improved BP performance over the previous comparative study, and that a simple variation of the ABC approach provides improved ABC performance too. With both improvements applied, we conclude that the ABC approach does perform very well on small problems, but the generalization performances achieved are only significantly better than standard BP on one out of six datasets, and the training times increase rapidly as the size of the problem grows.

We review here the logic of neuropsychological inference in the context of connectionist modelling, focusing on the inference from double dissociation to modularity of function. The results of an investigation into the effects of damage on a range of small artificial neural networks that have been trained to perform two distinct mappings (rules vs exceptions), suggest that a double dissociation is possible without modularity. However, when these studies are repeated using sufficiently larger and more distributed networks, which are presumably more psychologically and biologically relevant, double dissociations are not observed. Further analysis suggests that double dissociation between performance on rule-governed and exceptional items is only found when the contribution of individual units to the overall network performance is significant, and hence that such double dissociations are merely artefacts of scale. In large, fully distributed systems, a wide range of damage produces only a single dis...

After a brief review of the different types and causes of ambiguous training data and the problems of learning from such data, a class of multi-target models are presented which suggest that neural networks are even better at solving these problems than previously realized. They are able to learn which non-ambiguous subset of a larger ambiguous set of training data best captures any underlying regularities in that data and hence optimize generalization while minimizing the problems of overtraining. It is also shown how the deliberate generation of ambiguous training data can begin to solve some of the long-standing representational problems of mapping time sequences, such as the alignment problem for reading and spelling. The general ideas are illustrated throughout with the well- known problem of tex-to-phoneme conversion, and detailed results of a range of neural network simulations are presented.

Connectionist techniques are increasingly being used to model cognitive function with a view to providing extensions, elaborations or replacements of earlier “box and arrow” models. Networks of simplified processing units loosely based on real neurons are set up with architectures based on known physiology, trained to perform appropriately simplified versions of real tasks, and iteratively refined by checking their performance against humans. Such systems can still be linked together as in the old box and arrow models, with all the old explanations of patient data carrying through, but now we can examine the details of the degradation of the various components, and removing neurons or connections constitute natural analogues of real brain damage. We can also question the validity of the old assumptions of neuropsychological inference, and explore the possibility that processing is actually more distributed and interactive than the older models implied (Bullinaria, 2002). Here I shall outline what I consider to be the essential ideas.

A common problem in speech technology is the alignment of representations of text and phonemes, and the learning of a mapping between them that generalizes well to unseen inputs. The state-of-the-art technology appears to be symbolic rule-based systems, which is surprising given the number of neural network systems for text to phoneme mapping that have been developed over the years. This paper explores why that may be the case, and demonstrates that it is possible for neural networks to simultaneously perform text to phoneme alignment and mapping with performance levels at least comparable to the best existing systems.

Modularity is known to have benefits for neural systems and their evolution, and this paper aims to improve the evolutionary neural network algorithm EPNet to take advantage of those benefits. Neural networks exist with varying degrees of modularity ranging from pure modular networks characterized by disjoint partitions of hidden nodes with no communication between modules, to pure homogeneous networks with significant connections throughout. In between are apparently homogeneous networks that can be seen to have some degree of modularity if the hidden nodes are reorganized appropriately. In this paper, a modularity measure is presented and extended that can be applied to any neuron at any level in the network to provide a fine analysis of node partitioning. It also allows the rearrangement of nodes to create modules in homogeneous networks, and that is used to improve the EPNet algorithm to evolve modular neural networks. Experimental results on a simple classification task confirm that the new modular EPNet algorithm does indeed lead to more modular networks than the classical EPNet algorithm, without compromising the performance on the given task.

The Vehicle Routing Problem with Time Windows involves finding the lowest-cost set of routes to deliver goods to customers, which have service time windows, using a homogeneous fleet of vehicles with limited capacity. In this paper, we propose and analyze the performance of an improved multi-objective evolutionary algorithm, that simultaneously minimizes the number of routes, the total travel distance, and the delivery time. Empirical results indicate that the simultaneous minimization of all three objectives leads the algorithm to find similar or better results than any combination of only two objectives. These results, although not the best in all respects, are better in some aspects than all previously published approaches, and fully multi-objective comparisons show clear improvement over the popular NSGA-II algorithm.

The Vehicle Routing Problem can be seen as a fusion of two well known combinatorial problems, the Travelling Salesman Problem and Bin Packing Problem. It has several variants, the one with Time Windows being the case of study in this paper. Its main objective is to find the lowest-distance set of routes to deliver goods to customers, which have service time windows, using a fleet of identical vehicles with restricted capacity. We consider the simultaneous minimisation of the number of routes along with the total travel distance. Although previous research has considered evolutionary methods for solving this problem, none of them has concentrated on the similarity of solutions. We analyse here two methods to measure similarity, which are incorporated into an evolutionary algorithm to solve the multi-objective problem. We have applied this algorithm to a publicly available set of benchmark instances, and when these similarity measures are considered, our solutions are seen to be competitive or better than others previously published.

This work provides an analysis of using the evolutionary algorithm EPNet to create ensembles of artificial neural networks to solve a range of forecasting tasks. Several previous studies have tested the EPNet algorithm in the classification field, taking the best individuals to solve the problem and creating ensembles to improve the performance. But no studies have analyzed the behavior of the algorithm in detail for time series forecasting, nor used ensembles to try to improve the predictions. Thus, the aim of this work is to compare the ensemble approach, using two linear combination methods to calculate the output, against the best individual found. Since there are several parameters to adjust, experiments are set up to optimize them and improve the performance of the algorithm. The algorithm is tested on 21 time series of different behaviors. The experimental results show that, for time series forecasting, it is possible to improve the performance by using the ensemble method rather than using the best individual. This demonstrates that the information contained in the EPNet population is better than the information carried by any one individual.

This paper explores the idea that auto-teaching neural networks with evolved selfsupervision signals can lead to improved performance in dynamic environments where there is insufficient training data available within an individual’s lifetime. Results are presented from a series of artificial life experiments which investigate whether, when, and how this approach can lead to performance enhancements, in a simple problem domain that captures season dependent foraging behaviour.

The Vehicle Routing Problem with Time Windows is a complex combinatorial optimization problem which can be seen as a fusion of two well known sub-problems: the Travelling Salesman Problem and the Bin Packing Problem. Its main objective is to find the lowest-cost set of routes to deliver demand, using identical vehicles with limited capacity, to customers with fixed service time windows. In this paper, we consider the minimization of the number of routes and the total cost simultaneously. Although previous evolutionary studies have considered this problem, none of them has focused on the similarity of solutions in the population. We propose a method to measure route similarity and incorporate it into an evolutionary algorithm to solve the bi-objective VRPTW. We have applied this algorithm to a publicly available set of benchmark instances, resulting in solutions that are competitive or better than others previously published.

Alternative splicing is an important cellular process that allows the expression of a large number of unique cell-specific proteins from the same underlying strand of DNA, and thereby drastically increases the organism’s phenotypic plasticity. Its emergence is facilitated by the modular composition of genes into numerous semi-autonomous building blocks. In artificial evolution, such modular composition is usually unknown initially, but once learned may greatly increase the algorithm’s efficiency.

After a review of the purpose and biological background of Head Direction Cells, and existing models of them, we introduce an improved neural model that integrates visual flow information with the vestibular inputs found in earlier models. Simulation results using real video inputs are presented to explore and validate the new model, the evolutionary advantages for the proposed arrangement are noted, its utility for use in robot controllers is discussed, and suggestions are made for further work in this area.

In this paper we present a new nature-inspired variation operator for binary encodings in genetic algorithms (GAs). Our method, called implicit alternative splicing (iAS), is repeatedly applied to the individual encodings in the algorithm's population and inverts randomly chosen segments of decreasing size in a systematic fashion. Its goal is to determine the largest possible segment the inversion of which yields no loss in the encoding's quality. The application of iAS potentially produces a new encoding of equal or greater quality that is maximum possible Hamming distance from its source. This allows iAS to uphold the diversity of the population even if a minimal population size is chosen. This significantly improves the performance of an otherwise standard GA on a representative set of three different optimisation problems. Empirical results are compared and analysed and future work prospects are considered.

A novel evolutionary approach for the bin packing problem (BPP) is presented. A simple steady-state genetic algorithm is developed that produces results comparable to other approaches in the literature, without the need for any additional heuristics. The algorithm's design makes maximum use of the principle of natural selection to evolve valid solutions without the explicit need to verify constraint violations. Our algorithm is based upon a biologically inspired group encoding which allows for a modularisation of the search space in which individual sub-solutions may be assigned independent cost values. These values are subsequently utilised in a crossover event modelled on the theory of exon shuffling to produce a single offspring that inherits the most promising segments from its parents. The algorithm is tested on a set of hard benchmark problems and the results indicate that the method has a very high degree of accuracy and reliability compared to other approaches in the literature.

ExGA I, a previously presented genetic algorithm, successfully solved numerous instances of the multiple knapsack problem (MKS) by employing an adaptive repair function that made use of the algorithm's modular encoding. Here we present ExGA II, an extension of ExGA I that implements additional features which allow the algorithm to perform more reliably across a larger set of benchmark problems. In addition to some basic modifications of the algorithm's framework, more specific extensions include the use of a biased mutation operator and adaptive control sequences which are used to guide the repair procedure. The success rate of ExGA II is superior to its predecessor, and other algorithms in the literature, without an overall increase in the number of function evaluations required to reach the global optimum. In fact, the new algorithm exhibits a significant reduction in the number of function evaluations required for the largest problems investigated. We also address the computational cost of using a repair function and show that the algorithm remains highly competitive when this cost is accounted for.

A series of evolutionary neural network simulations are presented which explore the hypothesis that learning factors can result in the evolution of long periods of parental protection and late onset of maturity. By evolving populations of neural networks to learn quickly to perform well on simple classification tasks, it is shown that better learned performance is obtained if protection from competition is provided during the network's early learning period. Moreover, if the length of the protection period is allowed to evolve, it does result in the emergence of relatively long protection periods, even if there are other costs involved, such as individuals not being allowed to reproduce during their protection phase, and the parents suffering increased risk of dying while protecting their offspring.

Natural organisms have to deal efficiently with changing environments and are thus a great source of inspiration to solve dynamic problems in artificial domains. Dynamic optimisation has gained a lot of interest lately as many real world problems are indeed dynamic. In this paper, we look at post-transcriptional processes and alternative splicing in particular: Although these biochemical processes are gaining increasing attention from the genetics community, they remain relatively unexplored in evolutionary computation. We suggest a simple abstract encoding that allows one to construct multiple expressions from the same template supporting quick adaptation to changes in the cost surface. This encoding enables the system to find, control and reuse groups of building blocks that are being shared by different environments. Tests on a modified version of the dynamic knapsack problem show that it significantly outperforms the canonical genetic algorithm as well as simple implementations of random immigrants and hypermutations.

The idea of using evolutionary techniques to optimize the performance of neural networks is now widely used, but some approaches have been found to result in the evolution of risky learning strategies that lead to occasional instances of very poor performance. In this paper I shall present a series of simulations that explore the nature of those problems, and examine the extent to which one can use ensemble techniques to alleviate them.

It is well known that incremental learning can often be difficult for traditional neural network systems, due to newly learned information interfering with previously learned information. In this paper, we present simulation results which demonstrate how evolutionary computation techniques can be used to generate neural network incremental learners that exhibit improved performance over existing systems.

This paper presents the results of simulations of a new class of artificial neural network models of reading. Unlike previous models, they are not restricted to mono-syllabic words, require no complicated input-output representations such as Wickelfeatures and, although based on the NETtalk system of Sejnowski and Rosenberg (1987), require no pre-processing to align the letters and phonemes in the training data. The best cases are able to achieve 100% performance on the Seidenberg and McClelland (1989) training corpus, in excess of 90% on pronounceable nonwords and on damage exhibit symptoms similar to acquired surface dyslexia.

Attempts in the past to use evolutionary simulations to model the emergence of modular neural architectures has led to conflicting results. Here, we present some preliminary work on the use of computational embryogeny to model the evolution of neural architecture at a less coarse level of description. We believe that such an approach will lead to much more reliable and biologically realistic simulations of brain evolution.

Catastrophic forgetting is a well-known failing of many neural network systems whereby training on new patterns causes them to forget previously learned patterns. Humans have evolved mechanisms to minimize this problem, and in this paper we present our preliminary attempts to use simulated evolution to generate neural networks that suffer significantly less from catastrophic forgetting than traditionally formulated networks.

For autonomous neural network systems one usually needs fast learning and good generalization performance, and there is inevitably a trade-off between these two requirements. Using evolutionary techniques can generate high performance networks, but this often leads to unwanted side effects, such as occasional instances of very poor performance. This paper explores the problems that arise for traditional evolved neural networks using a range of evolutionary approaches, and shows how they can, to a large extent, be overcome by allowing the networks to evolve age dependent plasticities.

When training simple sigmoidal feed-forward neural networks on binary mappings using gradient descent algorithms with a sum-squared-error cost function, the learning algorithm often gets stuck with some outputs totally wrong. This is because the weight updates depend on the derivative of the output sigmoid which goes to zero as the output approaches maximal error. Common solutions to this problem include offsetting the output targets, offsetting the sigmoid derivatives, and using a different cost function. Comparisons are difficult because of the different optimal parameter settings for each case. In this paper I use an evolutionary approach to optimize and compare the different approaches.