Journal Articles
15.
Jordan J. Bird; Chloe M. Barnes; Luis J. Manso; Anikó Ekárt; Diego R. Faria
Fruit Quality and Defect Image Classification with Conditional GAN Data Augmentation Journal Article
In: Scientia Horticulturae, vol. 293, pp. 110684, 2022, ISSN: 0304-4238.
Abstract | Links | BibTeX | Tags: CNNs, Data Augmentation, Fruit Quality, GANs, Image Classification
@article{Bird2022FruitAugmentation,
title = {Fruit Quality and Defect Image Classification with Conditional GAN Data Augmentation},
author = {Jordan J. Bird and Chloe M. Barnes and Luis J. Manso and Anikó Ekárt and Diego R. Faria},
doi = {10.1016/j.scienta.2021.110684},
issn = {0304-4238},
year = {2022},
date = {2022-02-05},
urldate = {2022-01-01},
journal = {Scientia Horticulturae},
volume = {293},
pages = {110684},
abstract = {Contemporary Artificial Intelligence technologies allow for the employment of Computer Vision to discern good crops from bad, providing a step in the pipeline of selecting healthy fruit from undesirable fruit, such as those which are mouldy or damaged. State-of-the-art works in the field report high accuracy results on small datasets (<1000 images), which are not representative of the population regarding real-world usage. The goals of this study are to further enable real-world usage by improving generalisation with data augmentation as well as to reduce overfitting and energy usage through model pruning. In this work, we suggest a machine learning pipeline that combines the ideas of fine-tuning, transfer learning, and generative model-based training data augmentation towards improving fruit quality image classification. A linear network topology search is performed to tune a VGG16 lemon quality classification model using a publicly-available dataset of 2690 images. We find that appending a 4096 neuron fully connected layer to the convolutional layers leads to an image classification accuracy of 83.77%. We then train a Conditional Generative Adversarial Network on the training data for 2000 epochs, and it learns to generate relatively realistic images. Grad-CAM analysis of the model trained on real photographs shows that the synthetic images can exhibit classifiable characteristics such as shape, mould, and gangrene. A higher image classification accuracy of 88.75% is then attained by augmenting the training with synthetic images, arguing that Conditional Generative Adversarial Networks have the ability to produce new data to alleviate issues of data scarcity. Finally, model pruning is performed via polynomial decay, where we find that the Conditional GAN-augmented classification network can retain 81.16% classification accuracy when compressed to 50% of its original size.},
keywords = {CNNs, Data Augmentation, Fruit Quality, GANs, Image Classification},
pubstate = {published},
tppubtype = {article}
}
Contemporary Artificial Intelligence technologies allow for the employment of Computer Vision to discern good crops from bad, providing a step in the pipeline of selecting healthy fruit from undesirable fruit, such as those which are mouldy or damaged. State-of-the-art works in the field report high accuracy results on small datasets (<1000 images), which are not representative of the population regarding real-world usage. The goals of this study are to further enable real-world usage by improving generalisation with data augmentation as well as to reduce overfitting and energy usage through model pruning. In this work, we suggest a machine learning pipeline that combines the ideas of fine-tuning, transfer learning, and generative model-based training data augmentation towards improving fruit quality image classification. A linear network topology search is performed to tune a VGG16 lemon quality classification model using a publicly-available dataset of 2690 images. We find that appending a 4096 neuron fully connected layer to the convolutional layers leads to an image classification accuracy of 83.77%. We then train a Conditional Generative Adversarial Network on the training data for 2000 epochs, and it learns to generate relatively realistic images. Grad-CAM analysis of the model trained on real photographs shows that the synthetic images can exhibit classifiable characteristics such as shape, mould, and gangrene. A higher image classification accuracy of 88.75% is then attained by augmenting the training with synthetic images, arguing that Conditional Generative Adversarial Networks have the ability to produce new data to alleviate issues of data scarcity. Finally, model pruning is performed via polynomial decay, where we find that the Conditional GAN-augmented classification network can retain 81.16% classification accuracy when compressed to 50% of its original size.
14.
Chloe M. Barnes; Anikó Ekárt; Kai Olav Ellefsen; Kyrre Glette; Peter R. Lewis; Jim Tørresen
Behavioural Plasticity Can Help Evolving Agents in Dynamic Environments But at the Cost of Volatility Journal Article
In: ACM Transactions on Autonomous Adaptive Systems, vol. 15, no. 4, 2021.
Abstract | Links | BibTeX | Tags: Agent-Based Systems, ANNs, Artificial Life, Behavioural Plasticity, Evolutionary Algorithms, Evolutionary Volatility, Neuroevolution, Neuromodulation, River Crossing
@article{Barnes2021BehaviouralVolatility,
title = {Behavioural Plasticity Can Help Evolving Agents in Dynamic Environments But at the Cost of Volatility},
author = {Chloe M. Barnes and Anikó Ekárt and Kai Olav Ellefsen and Kyrre Glette and Peter R. Lewis and Jim Tørresen},
doi = {10.1145/3487918},
year = {2021},
date = {2021-12-20},
urldate = {2021-01-01},
journal = {ACM Transactions on Autonomous Adaptive Systems},
volume = {15},
number = {4},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
abstract = {Neural networks have been widely used in agent learning architectures; however, learnings for one task might nullify learnings for another. Behavioural plasticity enables humans and animals alike to respond to environmental changes without degrading learned knowledge; this can be achieved by regulating behaviour with neuromodulation—a biological process found in the brain. We demonstrate that by modulating activity-propagating signals, neurally trained agents evolving to solve tasks in dynamic environments that are prone to change can expect a significantly higher fitness than non-modulatory agents and also achieve their goals more often. Further, we show that while behavioural plasticity can help agents to achieve goals in these variable environments, this ability to overcome environmental changes with greater success comes at the cost of highly volatile evolution.},
keywords = {Agent-Based Systems, ANNs, Artificial Life, Behavioural Plasticity, Evolutionary Algorithms, Evolutionary Volatility, Neuroevolution, Neuromodulation, River Crossing},
pubstate = {published},
tppubtype = {article}
}
Neural networks have been widely used in agent learning architectures; however, learnings for one task might nullify learnings for another. Behavioural plasticity enables humans and animals alike to respond to environmental changes without degrading learned knowledge; this can be achieved by regulating behaviour with neuromodulation—a biological process found in the brain. We demonstrate that by modulating activity-propagating signals, neurally trained agents evolving to solve tasks in dynamic environments that are prone to change can expect a significantly higher fitness than non-modulatory agents and also achieve their goals more often. Further, we show that while behavioural plasticity can help agents to achieve goals in these variable environments, this ability to overcome environmental changes with greater success comes at the cost of highly volatile evolution.
13.
Chloe M. Barnes; Abida Ghouri; Peter R. Lewis
Explaining Evolutionary Agent-Based Models via Principled Simplification Journal Article
In: Artificial Life, vol. 27, no. 3, 2021.
Abstract | Links | BibTeX | Tags: Agent-Based Systems, ANNs, Artificial Life, Evolutionary Algorithms, Explainability, Neuroevolution, Principled Simplification, River Crossing
@article{Barnes2021ExplainingSimplification,
title = {Explaining Evolutionary Agent-Based Models via Principled Simplification},
author = {Chloe M. Barnes and Abida Ghouri and Peter R. Lewis},
doi = {10.1162/artl_a_00339},
year = {2021},
date = {2021-03-16},
urldate = {2021-01-01},
journal = {Artificial Life},
volume = {27},
number = {3},
publisher = {MIT Press},
abstract = {Understanding how evolutionary agents behave in complex environments is a challenging problem. Agents can be faced with complex fitness landscapes derived from multi-stage tasks, interaction with others, and limited environmental feedback. Agents that evolve to overcome these can sometimes access greater fitness, as a result of factors such as cooperation and tool use. However, it is often difficult to explain why evolutionary agents behave in certain ways, and what specific elements of the environment or task may influence the ability of evolution to find goal-achieving behaviours; even seemingly simple environments or tasks may contain features that affect agent evolution in unexpected ways. We explore principled simplification of evolutionary agent-based models, as a possible route to aiding their explainability. Using the River Crossing Task (RCT) as a case study, we draw on analysis in the Minimal River Crossing (RC-) Task testbed, which was designed to simplify the original task while keeping its key features. Using this method, we present new analysis concerning when agents evolve to successfully complete the RCT. We demonstrate that the RC- environment can be used to understand the effect that a cost to movement has on agent evolution, and that these findings can be generalised back to the original RCT. Then, we present new insight into the use of principled simplification in understanding evolutionary agents. We find evidence that behaviour dependent on features that survive simplification, such as problem structure, are amenable to prediction; while predicting behaviour dependent on features that are typically reduced in simplification, such as scale, can be invalid.},
keywords = {Agent-Based Systems, ANNs, Artificial Life, Evolutionary Algorithms, Explainability, Neuroevolution, Principled Simplification, River Crossing},
pubstate = {published},
tppubtype = {article}
}
Understanding how evolutionary agents behave in complex environments is a challenging problem. Agents can be faced with complex fitness landscapes derived from multi-stage tasks, interaction with others, and limited environmental feedback. Agents that evolve to overcome these can sometimes access greater fitness, as a result of factors such as cooperation and tool use. However, it is often difficult to explain why evolutionary agents behave in certain ways, and what specific elements of the environment or task may influence the ability of evolution to find goal-achieving behaviours; even seemingly simple environments or tasks may contain features that affect agent evolution in unexpected ways. We explore principled simplification of evolutionary agent-based models, as a possible route to aiding their explainability. Using the River Crossing Task (RCT) as a case study, we draw on analysis in the Minimal River Crossing (RC-) Task testbed, which was designed to simplify the original task while keeping its key features. Using this method, we present new analysis concerning when agents evolve to successfully complete the RCT. We demonstrate that the RC- environment can be used to understand the effect that a cost to movement has on agent evolution, and that these findings can be generalised back to the original RCT. Then, we present new insight into the use of principled simplification in understanding evolutionary agents. We find evidence that behaviour dependent on features that survive simplification, such as problem structure, are amenable to prediction; while predicting behaviour dependent on features that are typically reduced in simplification, such as scale, can be invalid.
12.
Chloe M. Barnes; Anikó Ekárt; Peter R. Lewis
Beyond Goal-Rationality: Traditional Action Can Reduce Volatility in Socially Situated Agents Journal Article
In: Future Generation Computer Systems, vol. 113, pp. 579–596, 2020.
Abstract | Links | BibTeX | Tags: Agent-Based Systems, ANNs, Artificial Life, Evolutionary Volatility, Neuroevolution, River Crossing, Social Action
@article{Barnes2020BeyondAgents,
title = {Beyond Goal-Rationality: Traditional Action Can Reduce Volatility in Socially Situated Agents},
author = {Chloe M. Barnes and Anikó Ekárt and Peter R. Lewis},
url = {https://www.sciencedirect.com/science/article/pii/S0167739X20304714},
doi = {10.1016/j.future.2020.07.033},
year = {2020},
date = {2020-12-01},
urldate = {2020-01-01},
journal = {Future Generation Computer Systems},
volume = {113},
pages = {579--596},
abstract = {Systems that pursue their own goals in shared environments can indirectly affect one another in unanticipated ways, such that the actions of other systems can interfere with goal-achievement. As humans have evolved to achieve goals despite interference from others in society, we thus endow socially situated agents with the capacity for social action as a means of mitigating interference in co-existing systems. We demonstrate that behavioural and evolutionary volatility caused by indirect interactions of goal-rational agents can be reduced by designing agents in a more socially-sensitive manner. We therefore challenge the assumption that designers of intelligent systems typically make, that goal-rationality is sufficient for achieving goals in shared environments.},
keywords = {Agent-Based Systems, ANNs, Artificial Life, Evolutionary Volatility, Neuroevolution, River Crossing, Social Action},
pubstate = {published},
tppubtype = {article}
}
Systems that pursue their own goals in shared environments can indirectly affect one another in unanticipated ways, such that the actions of other systems can interfere with goal-achievement. As humans have evolved to achieve goals despite interference from others in society, we thus endow socially situated agents with the capacity for social action as a means of mitigating interference in co-existing systems. We demonstrate that behavioural and evolutionary volatility caused by indirect interactions of goal-rational agents can be reduced by designing agents in a more socially-sensitive manner. We therefore challenge the assumption that designers of intelligent systems typically make, that goal-rationality is sufficient for achieving goals in shared environments.
11.
Jordan J. Bird; Chloe M. Barnes; Cristiano Premebida; Anikó Ekárt; Diego R. Faria
Country-level Pandemic Risk and Preparedness Classification based on COVID-19 Data: A Machine Learning Approach Journal Article
In: PLOS ONE, vol. 15, no. 10, pp. 1–20, 2020.
Abstract | Links | BibTeX | Tags: AI and Health, COVID-19, Machine Learning, Pandemic Risk Analysis
@article{Bird2020Country-levelApproach,
title = {Country-level Pandemic Risk and Preparedness Classification based on COVID-19 Data: A Machine Learning Approach},
author = {Jordan J. Bird and Chloe M. Barnes and Cristiano Premebida and Anikó Ekárt and Diego R. Faria},
doi = {10.1371/journal.pone.0241332},
year = {2020},
date = {2020-10-10},
urldate = {2020-01-01},
journal = {PLOS ONE},
volume = {15},
number = {10},
pages = {1--20},
publisher = {Public Library of Science},
abstract = {In this work we present a three-stage Machine Learning strategy to country-level risk classification based on countries that are reporting COVID-19 information. A K% binning discretisation (K = 25) is used to create four risk groups of countries based on the risk of transmission (coronavirus cases per million population), risk of mortality (coronavirus deaths per million population), and risk of inability to test (coronavirus tests per million population). The four risk groups produced by K% binning are labelled as ‘low’, ‘medium-low’, ‘medium-high’, and ‘high’. Coronavirus-related data are then removed and the attributes for prediction of the three types of risk are given as the geopolitical and demographic data describing each country. Thus, the calculation of class label is based on coronavirus data but the input attributes are country-level information regardless of coronavirus data. The three four-class classification problems are then explored and benchmarked through leave-one-country-out cross validation to find the strongest model, producing a Stack of Gradient Boosting and Decision Tree algorithms for risk of transmission, a Stack of Support Vector Machine and Extra Trees for risk of mortality, and a Gradient Boosting algorithm for the risk of inability to test. It is noted that high risk for inability to test is often coupled with low risks for transmission and mortality, therefore the risk of inability to test should be interpreted first, before consideration is given to the predicted transmission and mortality risks. Finally, the approach is applied to more recent risk levels to data from September 2020 and weaker results are noted due to the growth of international collaboration detracting useful knowledge from country-level attributes which suggests that similar machine learning approaches are more useful prior to situations later unfolding.},
keywords = {AI and Health, COVID-19, Machine Learning, Pandemic Risk Analysis},
pubstate = {published},
tppubtype = {article}
}
In this work we present a three-stage Machine Learning strategy to country-level risk classification based on countries that are reporting COVID-19 information. A K% binning discretisation (K = 25) is used to create four risk groups of countries based on the risk of transmission (coronavirus cases per million population), risk of mortality (coronavirus deaths per million population), and risk of inability to test (coronavirus tests per million population). The four risk groups produced by K% binning are labelled as ‘low’, ‘medium-low’, ‘medium-high’, and ‘high’. Coronavirus-related data are then removed and the attributes for prediction of the three types of risk are given as the geopolitical and demographic data describing each country. Thus, the calculation of class label is based on coronavirus data but the input attributes are country-level information regardless of coronavirus data. The three four-class classification problems are then explored and benchmarked through leave-one-country-out cross validation to find the strongest model, producing a Stack of Gradient Boosting and Decision Tree algorithms for risk of transmission, a Stack of Support Vector Machine and Extra Trees for risk of mortality, and a Gradient Boosting algorithm for the risk of inability to test. It is noted that high risk for inability to test is often coupled with low risks for transmission and mortality, therefore the risk of inability to test should be interpreted first, before consideration is given to the predicted transmission and mortality risks. Finally, the approach is applied to more recent risk levels to data from September 2020 and weaker results are noted due to the growth of international collaboration detracting useful knowledge from country-level attributes which suggests that similar machine learning approaches are more useful prior to situations later unfolding.
Inproceedings
10.
Sam G. Robinson; Chloe M. Barnes; Peter R. Lewis
Centralised and Decentralised Control of Video Game Agents Inproceedings
In: Advances in Computational Intelligence Systems. UKCI 2021, pp. 108–120, Springer, 2022.
Abstract | Links | BibTeX | Tags: ANNs, Game Playing, Network Comparisons, Neuroevolution
@inproceedings{Robinson2021CentralisedAgents,
title = {Centralised and Decentralised Control of Video Game Agents},
author = {Sam G. Robinson and Chloe M. Barnes and Peter R. Lewis},
doi = {10.1007/978-3-030-87094-2_10},
year = {2022},
date = {2022-09-08},
urldate = {2022-01-01},
booktitle = {Advances in Computational Intelligence Systems. UKCI 2021},
volume = {1409},
pages = {108--120},
publisher = {Springer},
series = {Advances in Intelligent Systems and Computing},
abstract = {In this paper, the game of partially observable Ms. Pacman is used as a sandbox to evaluate Artificial Neural Networks (ANNs) that control multiple opponents (i.e. the ghosts). Comparisons between one central ANN that controls all ghosts, and multiple distinct ANNs, each controlling one ghost, are made. The NEAT algorithm is employed to evolve the ANNs. We find that chasing Ms. Pacman and exploring the map are both harder behaviours to learn for a centralised controller than for decentralised control. Further, both centralised and decentralised approaches produce vastly different behaviours for exploring the map. Novel techniques for comparing networks are also explored.},
keywords = {ANNs, Game Playing, Network Comparisons, Neuroevolution},
pubstate = {published},
tppubtype = {inproceedings}
}
In this paper, the game of partially observable Ms. Pacman is used as a sandbox to evaluate Artificial Neural Networks (ANNs) that control multiple opponents (i.e. the ghosts). Comparisons between one central ANN that controls all ghosts, and multiple distinct ANNs, each controlling one ghost, are made. The NEAT algorithm is employed to evolve the ANNs. We find that chasing Ms. Pacman and exploring the map are both harder behaviours to learn for a centralised controller than for decentralised control. Further, both centralised and decentralised approaches produce vastly different behaviours for exploring the map. Novel techniques for comparing networks are also explored.
9.
Chloe M. Barnes
A Study on Psychometric Assessment Data for Autonomous Dementia Detection Inproceedings
In: Proceedings of the 15th International Conference on PErvasive Technologies Related to Assistive Environments, pp. 383–389, Association for Computing Machinery, Corfu, Greece, 2022.
Abstract | Links | BibTeX | Tags: AI and Health, Dementia Classification, Machine Learning, Multi-Modal Data
@inproceedings{10.1145/3529190.3534726,
title = {A Study on Psychometric Assessment Data for Autonomous Dementia Detection},
author = {Chloe M. Barnes},
url = {https://doi.org/10.1145/3529190.3534726},
doi = {10.1145/3529190.3534726},
year = {2022},
date = {2022-06-29},
urldate = {2022-01-01},
booktitle = {Proceedings of the 15th International Conference on PErvasive Technologies Related to Assistive Environments},
pages = {383--389},
publisher = {Association for Computing Machinery},
address = {Corfu, Greece},
series = {PETRA '22},
abstract = {Dementia and Alzheimer’s disease are characterised by cognitive decline, and diagnoses are predicted to rise due to an ageing population. Psychometric assessments are widely used by clinicians to inform the diagnosis of dementia, however these may not be as accurate or accessible for patients with lower levels of literacy or socioeconomic status. This study explores how machine learning models can detect dementia when trained on combinations of attributes from multi-modal datasets containing psychometric and Magnetic Resonance Imaging (MRI) data. When psychometric testing is not available, results show that the Random Forest classifier achieves a balanced accuracy, sensitivity and specificity of 84.75%, 79.10%, and 90.41% respectively before the dataset was standardised, and 84.34%, 78.27%, and 90.41% after – outperforming identical models trained on data from a single psychometric test.},
keywords = {AI and Health, Dementia Classification, Machine Learning, Multi-Modal Data},
pubstate = {published},
tppubtype = {inproceedings}
}
Dementia and Alzheimer’s disease are characterised by cognitive decline, and diagnoses are predicted to rise due to an ageing population. Psychometric assessments are widely used by clinicians to inform the diagnosis of dementia, however these may not be as accurate or accessible for patients with lower levels of literacy or socioeconomic status. This study explores how machine learning models can detect dementia when trained on combinations of attributes from multi-modal datasets containing psychometric and Magnetic Resonance Imaging (MRI) data. When psychometric testing is not available, results show that the Random Forest classifier achieves a balanced accuracy, sensitivity and specificity of 84.75%, 79.10%, and 90.41% respectively before the dataset was standardised, and 84.34%, 78.27%, and 90.41% after – outperforming identical models trained on data from a single psychometric test.
8.
Chloe M. Barnes; Anikó Ekárt; Kai Olav Ellefsen; Kyrre Glette; Peter R. Lewis; Jim Tørresen
Evolving Neuromodulated Controllers in Variable Environments Inproceedings
In: Proceedings of the IEEE 2nd International Conference on Autonomic Computing and Self-Organizing Systems (ACSOS), pp. 164–169, IEEE, 2021.
Abstract | Links | BibTeX | Tags: Agent-Based Systems, ANNs, Artificial Life, Behavioural Plasticity, Cooperation, Environmental Variability, Evolutionary Algorithms, Neuroevolution, Neuromodulation, River Crossing
@inproceedings{Barnes2021EvolvingEnvironments,
title = {Evolving Neuromodulated Controllers in Variable Environments},
author = {Chloe M. Barnes and Anikó Ekárt and Kai Olav Ellefsen and Kyrre Glette and Peter R. Lewis and Jim Tørresen},
doi = {10.1109/ACSOS52086.2021.00037},
year = {2021},
date = {2021-09-01},
urldate = {2021-01-01},
booktitle = {Proceedings of the IEEE 2nd International Conference on Autonomic Computing and Self-Organizing Systems (ACSOS)},
pages = {164--169},
publisher = {IEEE},
abstract = {Modern technical systems are increasingly composed of heterogeneous components that are situated in variable environments. In nature, organisms can temporarily adapt their behaviour to novel stimuli with behavioural plasticity; this can be achieved with neuromodulation, a biological process that modulates synaptic activity in the brain. We explore how neuromodulation affects goal-achievement in evolved neural controllers for artificial agents in variable environments. As variability can arise from the actions of others, we show that the benefit of plasticity can increase with variability, as agents can temporarily change their phenotype within their lifetime. Further, we show that cooperation can emerge between plastic agents that cannot perceive one another in highly variable environments.},
keywords = {Agent-Based Systems, ANNs, Artificial Life, Behavioural Plasticity, Cooperation, Environmental Variability, Evolutionary Algorithms, Neuroevolution, Neuromodulation, River Crossing},
pubstate = {published},
tppubtype = {inproceedings}
}
Modern technical systems are increasingly composed of heterogeneous components that are situated in variable environments. In nature, organisms can temporarily adapt their behaviour to novel stimuli with behavioural plasticity; this can be achieved with neuromodulation, a biological process that modulates synaptic activity in the brain. We explore how neuromodulation affects goal-achievement in evolved neural controllers for artificial agents in variable environments. As variability can arise from the actions of others, we show that the benefit of plasticity can increase with variability, as agents can temporarily change their phenotype within their lifetime. Further, we show that cooperation can emerge between plastic agents that cannot perceive one another in highly variable environments.
7.
Chloe M. Barnes; Anikó Ekárt; Kai Olav Ellefsen; Kyrre Glette; Peter R. Lewis; Jim Tørresen
Coevolutionary Learning of Neuromodulated Controllers for Multi-Stage and Gamified Tasks Inproceedings
In: Proceedings of the IEEE 1st International Conference on Autonomic Computing and Self-Organizing Systems (ACSOS), pp. 129–138, IEEE, 2020.
Abstract | Links | BibTeX | Tags: Agent-Based Systems, ANNs, Artificial Life, Behavioural Plasticity, Evolutionary Algorithms, Neuroevolution, Neuromodulation, River Crossing
@inproceedings{Barnes2020CoevolutionaryTasks,
title = {Coevolutionary Learning of Neuromodulated Controllers for Multi-Stage and Gamified Tasks},
author = {Chloe M. Barnes and Anikó Ekárt and Kai Olav Ellefsen and Kyrre Glette and Peter R. Lewis and Jim Tørresen},
url = {https://ieeexplore.ieee.org/document/9196458},
doi = {10.1109/ACSOS49614.2020.00034},
year = {2020},
date = {2020-08-17},
urldate = {2020-01-01},
booktitle = {Proceedings of the IEEE 1st International Conference on Autonomic Computing and Self-Organizing Systems (ACSOS)},
pages = {129--138},
publisher = {IEEE},
abstract = {Neural networks have been widely used in agent learning architectures; however, learning multiple context dependent tasks simultaneously or sequentially is problematic when using them. Behavioural plasticity enables humans and animals alike to respond to changes in context and environmental stimuli, without degrading learnt knowledge; this can be achieved by regulating behaviour with neuromodulation - a biological process found in the brain. We demonstrate that modulating activity-propagating signals when evolving neural networks enables agents to learn context-dependent and multi-stage tasks more easily. Further, we show that this benefit is preserved when agents occupy an environment shared with other neuromodulated agents. Additionally we show that neuromodulation helps agents that have evolved alone to adapt to changes in environmental stimuli when they continue to evolve in a shared environment.},
keywords = {Agent-Based Systems, ANNs, Artificial Life, Behavioural Plasticity, Evolutionary Algorithms, Neuroevolution, Neuromodulation, River Crossing},
pubstate = {published},
tppubtype = {inproceedings}
}
Neural networks have been widely used in agent learning architectures; however, learning multiple context dependent tasks simultaneously or sequentially is problematic when using them. Behavioural plasticity enables humans and animals alike to respond to changes in context and environmental stimuli, without degrading learnt knowledge; this can be achieved by regulating behaviour with neuromodulation - a biological process found in the brain. We demonstrate that modulating activity-propagating signals when evolving neural networks enables agents to learn context-dependent and multi-stage tasks more easily. Further, we show that this benefit is preserved when agents occupy an environment shared with other neuromodulated agents. Additionally we show that neuromodulation helps agents that have evolved alone to adapt to changes in environmental stimuli when they continue to evolve in a shared environment.
6.
Abida Ghouri; Chloe M. Barnes; Peter R. Lewis
A Minimal River Crossing Task to Aid the Explainability of Evolutionary Agents Inproceedings
In: ALIFE 2020: The 2020 Conference on Artificial Life, pp. 36–43, 2020.
Abstract | Links | BibTeX | Tags: Agent-Based Systems, ANNs, Artificial Life, Evolutionary Algorithms, Explainability, Neuroevolution, Principled Simplification, River Crossing
@inproceedings{Ghouri2020AAgents,
title = {A Minimal River Crossing Task to Aid the Explainability of Evolutionary Agents},
author = {Abida Ghouri and Chloe M. Barnes and Peter R. Lewis},
url = {https://www.mitpressjournals.org/doi/abs/10.1162/isal_a_00347},
doi = {10.1162/isal_a_00347},
year = {2020},
date = {2020-07-01},
urldate = {2020-01-01},
booktitle = {ALIFE 2020: The 2020 Conference on Artificial Life},
pages = {36--43},
abstract = {n
Evolving agents to learn how to solve complex, multi-stage tasks to achieve a goal is a challenging problem. Problems such as the River Crossing Task are used to explore how these agents evolve and what they learn, but it is still often difficult to explain why agents behave in the way they do. We present the Minimal River Crossing (RC-) Task testbed, designed to reduce the complexity of the original River Crossing Task while keeping its essential components, such that the fundamental learning challenges it presents can be understood in more detail. Specifically to illustrate this, we demonstrate that the RC-environment can be used to investigate the effect that a cost to movement has on agent evolution and learning, and more importantly that the findings obtained as a result can be generalised back to the original River Crossing Task.},
keywords = {Agent-Based Systems, ANNs, Artificial Life, Evolutionary Algorithms, Explainability, Neuroevolution, Principled Simplification, River Crossing},
pubstate = {published},
tppubtype = {inproceedings}
}
n
Evolving agents to learn how to solve complex, multi-stage tasks to achieve a goal is a challenging problem. Problems such as the River Crossing Task are used to explore how these agents evolve and what they learn, but it is still often difficult to explain why agents behave in the way they do. We present the Minimal River Crossing (RC-) Task testbed, designed to reduce the complexity of the original River Crossing Task while keeping its essential components, such that the fundamental learning challenges it presents can be understood in more detail. Specifically to illustrate this, we demonstrate that the RC-environment can be used to investigate the effect that a cost to movement has on agent evolution and learning, and more importantly that the findings obtained as a result can be generalised back to the original River Crossing Task.
Evolving agents to learn how to solve complex, multi-stage tasks to achieve a goal is a challenging problem. Problems such as the River Crossing Task are used to explore how these agents evolve and what they learn, but it is still often difficult to explain why agents behave in the way they do. We present the Minimal River Crossing (RC-) Task testbed, designed to reduce the complexity of the original River Crossing Task while keeping its essential components, such that the fundamental learning challenges it presents can be understood in more detail. Specifically to illustrate this, we demonstrate that the RC-environment can be used to investigate the effect that a cost to movement has on agent evolution and learning, and more importantly that the findings obtained as a result can be generalised back to the original River Crossing Task.
5.
Chloe M. Barnes; Kirstie Bellman; Jean Botev; Ada Diaconescu; Lukas Esterle; Christian Gruhl; Chris Landauer; Peter R. Lewis; Phyllis R. Nelson; Anthony Stein; Christopher Stewart; Sven Tomforde
CHARIOT -- Towards a Continuous High-Level Adaptive Runtime Integration Testbed Inproceedings
In: Proceedings of the IEEE 4th International Workshops on Foundations and Applications of Self* Systems (FAS*W), pp. 52–55, IEEE, 2019.
Abstract | Links | BibTeX | Tags: Adaptive Systems, Runtime Systems Integration
@inproceedings{Barnes2019CHARIOTTestbed,
title = {CHARIOT -- Towards a Continuous High-Level Adaptive Runtime Integration Testbed},
author = {Chloe M. Barnes and Kirstie Bellman and Jean Botev and Ada Diaconescu and Lukas Esterle and Christian Gruhl and Chris Landauer and Peter R. Lewis and Phyllis R. Nelson and Anthony Stein and Christopher Stewart and Sven Tomforde},
url = {https://ieeexplore.ieee.org/document/8791947},
doi = {10.1109/FAS-W.2019.00026},
year = {2019},
date = {2019-06-19},
urldate = {2019-01-01},
booktitle = {Proceedings of the IEEE 4th International Workshops on Foundations and Applications of Self* Systems (FAS*W)},
pages = {52--55},
publisher = {IEEE},
abstract = {Integrated networked systems sense a common environment, learn to navigate the environment and share their experiences. Sharing experiences simplifies learning, reducing costly trial and error in complex environments. However, integration produces dependencies that make constituent systems less robust to failures, unexpected outputs and performance anomalies. Even with APIs and reflective, self-aware techniques, system integration still requires expert programming and tuning. Self-integrating systems proposed in recent research automate integration, but can be challenging to validate at scale. We therefore propose CHARIOT, a common test environment to allow for different approaches and systems to be deployed, assessed and compared on a shared platform for the development of self-integrating systems. In this paper, we discuss the underlying requirements and challenges, potential metrics, and a system metamodel to accommodate these.},
keywords = {Adaptive Systems, Runtime Systems Integration},
pubstate = {published},
tppubtype = {inproceedings}
}
Integrated networked systems sense a common environment, learn to navigate the environment and share their experiences. Sharing experiences simplifies learning, reducing costly trial and error in complex environments. However, integration produces dependencies that make constituent systems less robust to failures, unexpected outputs and performance anomalies. Even with APIs and reflective, self-aware techniques, system integration still requires expert programming and tuning. Self-integrating systems proposed in recent research automate integration, but can be challenging to validate at scale. We therefore propose CHARIOT, a common test environment to allow for different approaches and systems to be deployed, assessed and compared on a shared platform for the development of self-integrating systems. In this paper, we discuss the underlying requirements and challenges, potential metrics, and a system metamodel to accommodate these.
4.
Chloe M. Barnes; Anikó Ekárt; Peter R. Lewis
Social Action in Socially Situated Agents Inproceedings
In: Proceedings of the IEEE 13th International Conference on Self-Adaptive and Self-Organizing Systems (SASO), pp. 97–106, IEEE, 2019.
Abstract | Links | BibTeX | Tags: Agent-Based Systems, ANNs, Artificial Life, Evolutionary Algorithms, Neuroevolution, River Crossing, Social Action
@inproceedings{Barnes2019saso,
title = {Social Action in Socially Situated Agents},
author = {Chloe M. Barnes and Anikó Ekárt and Peter R. Lewis},
url = {https://ieeexplore.ieee.org/document/8780530},
doi = {10.1109/SASO.2019.00021},
year = {2019},
date = {2019-06-19},
urldate = {2019-01-01},
booktitle = {Proceedings of the IEEE 13th International Conference on Self-Adaptive and Self-Organizing Systems (SASO)},
pages = {97--106},
publisher = {IEEE},
abstract = {Two systems pursuing their own goals in a shared world can interact in ways that are not so explicit - such that the presence of another system alone can interfere with how one is able to achieve its own goals. Drawing inspiration from human psychology and the theory of social action, we propose the notion of employing social action in socially situated agents as a means of alleviating interference in interacting systems. Here we demonstrate that these specific issues of behavioural and evolutionary instability caused by the unintended consequences of interactions can be addressed with agents capable of a fusion of goal-rationality and traditional action, resulting in a stable society capable of achieving goals during the course of evolution.},
keywords = {Agent-Based Systems, ANNs, Artificial Life, Evolutionary Algorithms, Neuroevolution, River Crossing, Social Action},
pubstate = {published},
tppubtype = {inproceedings}
}
Two systems pursuing their own goals in a shared world can interact in ways that are not so explicit - such that the presence of another system alone can interfere with how one is able to achieve its own goals. Drawing inspiration from human psychology and the theory of social action, we propose the notion of employing social action in socially situated agents as a means of alleviating interference in interacting systems. Here we demonstrate that these specific issues of behavioural and evolutionary instability caused by the unintended consequences of interactions can be addressed with agents capable of a fusion of goal-rationality and traditional action, resulting in a stable society capable of achieving goals during the course of evolution.
3.
Chloe M. Barnes; Lukas Esterle; John N. A. Brown
``When You Believe in Things That You Don't Understand Inproceedings
In: Proceedings of the IEEE 4th International Workshops on Foundations and Applications of Self* Systems (FAS*W), pp. 28–31, IEEE, 2019.
Abstract | Links | BibTeX | Tags: Computational Superstition, Networked Self-Awareness, Runtime Systems Integration, Self-Awareness
@inproceedings{Barnes2019SISSY,
title = {``When You Believe in Things That You Don't Understand},
author = {Chloe M. Barnes and Lukas Esterle and John N. A. Brown},
url = {https://ieeexplore.ieee.org/abstract/document/8791979},
doi = {10.1109/FAS-W.2019.00020},
year = {2019},
date = {2019-06-16},
urldate = {2019-01-01},
booktitle = {Proceedings of the IEEE 4th International Workshops on Foundations and Applications of Self* Systems (FAS*W)},
pages = {28--31},
publisher = {IEEE},
abstract = {Humans experiencing unexpected feedback to certain actions which they are not able to explain, might develop superstitious behaviour. In this paper, we discuss that similar behaviour might also occur in engineered systems. We provide a thought-experiment regarding such behaviour in computational systems. In this paper, we propose a first step towards improved runtime systems integration based on a the ability to become aware of previously-unknown others and their actions, as described in networked self-awareness.},
keywords = {Computational Superstition, Networked Self-Awareness, Runtime Systems Integration, Self-Awareness},
pubstate = {published},
tppubtype = {inproceedings}
}
Humans experiencing unexpected feedback to certain actions which they are not able to explain, might develop superstitious behaviour. In this paper, we discuss that similar behaviour might also occur in engineered systems. We provide a thought-experiment regarding such behaviour in computational systems. In this paper, we propose a first step towards improved runtime systems integration based on a the ability to become aware of previously-unknown others and their actions, as described in networked self-awareness.
PhD Theses
2.
Chloe M. Barnes
Interference and Volatility in Evolutionary Agent-Based Systems PhD Thesis
Aston University, 2021.
Abstract | BibTeX | Tags: Agent-Based Systems, ANNs, Artificial Life, Behavioural Plasticity, Cooperation, Environmental Variability, Evolutionary Algorithms, Evolutionary Volatility, Network Comparisons, Neuroevolution, Neuromodulation, River Crossing
@phdthesis{Barnes2021thesis,
title = {Interference and Volatility in Evolutionary Agent-Based Systems},
author = {Chloe M. Barnes},
year = {2021},
date = {2021-09-20},
urldate = {2021-01-01},
school = {Aston University},
abstract = {Agents that exist and pursue individual goals in shared environments can indirectly affect one another in unanticipated ways, such that the actions of others in the environment can interfere with the ability to achieve goals. Despite this, the impact that these unintended interactions and interference can have on agents is not currently well understood. This is problematic as these goal-oriented agents are increasingly situated in complex sociotechnical systems, that are composed of many actors that are heterogeneous in nature. The primary aim of this thesis is to explore the effect that indirect interference from others has on evolution and goal-achieving behaviour in agent-based systems. More specifically, this is investigated in the context of agents that do not possess the ability to perceive or learn about others within the environment, as information about others may not be readily available at runtime, or there may be a distinct lack of capacity to obtain such information. By conducting three experimental studies, it is established that evolutionary volatility is a consequence of indirect interactions between goal-oriented agents in a shared environment, and that these consequences can be mitigated by designing more socially-sensitive agents. Specifically, agents that employ social action are demonstrated to reduce the evolutionary volatility experienced by goal-oriented agents, without aecting the tness received. Additionally, behavioural plasticity achieved via neuromodulation is shown to allow coexisting agents to achieve their goals more often with less evolutionary volatility in highly variable environments. While sufficient approaches to mitigate interference include learning about or modelling others, or for agents to be explicitly designed to identify interference to mitigate its consequences, this thesis demonstrates that these are not necessary. Instead, more socially-sensitive agents are shown to be capable of achieving their goals and mitigating interference without this knowledge of others, simply by shifting the focus from goal-oriented actions to more socially-oriented behaviour.},
keywords = {Agent-Based Systems, ANNs, Artificial Life, Behavioural Plasticity, Cooperation, Environmental Variability, Evolutionary Algorithms, Evolutionary Volatility, Network Comparisons, Neuroevolution, Neuromodulation, River Crossing},
pubstate = {published},
tppubtype = {phdthesis}
}
Agents that exist and pursue individual goals in shared environments can indirectly affect one another in unanticipated ways, such that the actions of others in the environment can interfere with the ability to achieve goals. Despite this, the impact that these unintended interactions and interference can have on agents is not currently well understood. This is problematic as these goal-oriented agents are increasingly situated in complex sociotechnical systems, that are composed of many actors that are heterogeneous in nature. The primary aim of this thesis is to explore the effect that indirect interference from others has on evolution and goal-achieving behaviour in agent-based systems. More specifically, this is investigated in the context of agents that do not possess the ability to perceive or learn about others within the environment, as information about others may not be readily available at runtime, or there may be a distinct lack of capacity to obtain such information. By conducting three experimental studies, it is established that evolutionary volatility is a consequence of indirect interactions between goal-oriented agents in a shared environment, and that these consequences can be mitigated by designing more socially-sensitive agents. Specifically, agents that employ social action are demonstrated to reduce the evolutionary volatility experienced by goal-oriented agents, without aecting the tness received. Additionally, behavioural plasticity achieved via neuromodulation is shown to allow coexisting agents to achieve their goals more often with less evolutionary volatility in highly variable environments. While sufficient approaches to mitigate interference include learning about or modelling others, or for agents to be explicitly designed to identify interference to mitigate its consequences, this thesis demonstrates that these are not necessary. Instead, more socially-sensitive agents are shown to be capable of achieving their goals and mitigating interference without this knowledge of others, simply by shifting the focus from goal-oriented actions to more socially-oriented behaviour.
Presentations
1.
Chloe M. Barnes; Anikó Ekárt; Kai Olav Ellefsen; Kyrre Glette; Peter R. Lewis; Jim Tørresen
Overcoming Dynamicity with Plasticity: Neuromodulation for Lifelike Systems Presentation
LIFELIKE Systems Workshop, 2022, 20.07.2022.
Abstract | Links | BibTeX | Tags: Agent-Based Systems, ANNs, Artificial Life, Behavioural Plasticity, Environmental Variability, Evolutionary Algorithms, Evolutionary Volatility, Neuroevolution, Neuromodulation, River Crossing
@misc{Barnes2022OvercomingSystems,
title = {Overcoming Dynamicity with Plasticity: Neuromodulation for Lifelike Systems},
author = {Chloe M. Barnes and Anikó Ekárt and Kai Olav Ellefsen and Kyrre Glette and Peter R. Lewis and Jim Tørresen},
url = {https://www.organic-computing.de/wp-content/uploads/2022/08/LIFELIKE2022_Barnes-et-al_Neuromodulation.pdf},
year = {2022},
date = {2022-07-20},
urldate = {2022-01-01},
abstract = {Natural beings are often situated in dynamic and unpredictable environments, and have evolved to use mechanisms such as neuromodulation -- the ability to change behaviour via changes to synaptic activity in the brain -- to adapt their behaviour over time to survive. The ability to change behaviour in this way is referred to as behavioural plasticity'. In this extended abstract, we summarise the findings from an exploration of how plasticity can affect how artificial agents evolve when solving tasks of different complexity, and when evolving in dynamic and unpredictable environments.},
howpublished = {LIFELIKE Systems Workshop, 2022},
keywords = {Agent-Based Systems, ANNs, Artificial Life, Behavioural Plasticity, Environmental Variability, Evolutionary Algorithms, Evolutionary Volatility, Neuroevolution, Neuromodulation, River Crossing},
pubstate = {published},
tppubtype = {presentation}
}
Natural beings are often situated in dynamic and unpredictable environments, and have evolved to use mechanisms such as neuromodulation -- the ability to change behaviour via changes to synaptic activity in the brain -- to adapt their behaviour over time to survive. The ability to change behaviour in this way is referred to as behavioural plasticity'. In this extended abstract, we summarise the findings from an exploration of how plasticity can affect how artificial agents evolve when solving tasks of different complexity, and when evolving in dynamic and unpredictable environments.
